Thursday, December 28, 2006

2007 -- A Year Of Promise

Tomorrow will be a new year. A lot of people will be looking for the Q4 06 results and making predictions for Intel and AMD in 2007. However, given what I've heard in the last couple of months it will be difficult to sort the wheat from the chafe. I'd like to look at where things are now and where they are going.

Reality has been a bit at odds with appearances in 2006. All of the fanfare has been from Intel. Revision F from AMD was a minor announcement as was 4x4. In contrast, Intel started the ad campaign blaring back in Q2 long before Core 2 Duo was actually available. And, it hasn't stopped. We have seen review after review of Woodcrest and Conroe and Kentsfield and Clovertown. Each review has been so hopeful and practically giddy with all of the incredible speed of C2D over K8 as well as lower power draw. Add to that, all of Intel's production capacity and AMD's capacity limitations and many Intel proponents were ready for announcements that Intel had taken back significant share from AMD. Except, it didn't happen.

The fourth quarter of 2005 was the best quarter that Intel has ever had in terms of processor sales. Yet, this didn't stop AMD from taking 4% volume share from Intel. I was a bit cautious about this in the first quarter of 2006 because I thought the gain could have been temporary and the volume share could have switched back. However, AMD not only held onto its 4% but managed to add another 1.5% giving it a total gain of 5.5% by Q3. Intel has gained none of this back nor has Intel gained any microprocessor revenue share. I'm guessing that some are waiting for the Q4 reports to see if Intel has gained anything. However, with AMD's capacity constraints Intel could have temporary gains if the market volume surges more than AMD can match. This wouldn't be a real gain because the % share would fall back again in Q1 07 when the volume drops. The best time to see volume share will be the second quarter when volume is lowest. Sales at that point should reflect a true preference in hardware for the market unless the total revenue falls. With falling revenue, share increases by AMD may not be accurate which is why I wasn't sure about AMD's volume share increases until Q3 of 2006. However, if the market stays up then Q2 should be accurate.

That Intel has not gained share is simply not believable to some. They point out that C2D is better and therefore Intel should gain. The problem with this idea is several things. First of all, Intel started producing C2D in mid year and then has ramped to 25% of production by today. This puts C2D at only 6.25% of production in Q3 and 18.75% of production in Q4. So, C2D accounted for only 12.5% of Intel's production for 2006. This has simply not been enough to have much effect on share. AMD will leave 2006 with the best year it has ever had with twice the processor revenue that it had 2 years ago. In contrast, Intel will leave 2006 with about the same processor revenue as 2 years ago. I think anyone with common sense can see that AMD has done much better than Intel in 2006.

Seemingly to fix what some must perceive as a devine right by Intel to take back share, some have invented paper gains for Intel by including the X86 sales from VIA with the volume figures. With one of their processors being end of life, VIA dumped these on the market. If we include these processors in the numbers then this lowers Intel's Q2 share even further which then makes the Q3 numbers appear to be increasing because the VIA share is gone again. However, these are only paper gains and don't reflect any actual increase by Intel. If we look at the share between AMD and Intel only then Intel has lost more share since Q4 05 and there has been no gain for Intel in Q3. Those who stoically ignore such attempts or perhaps who find the small paper gains to be insufficient then turn to early 2007 with great hope for share gains by Intel. The basic argument is that Intel is still ahead with C2D and with its higher percentage of production they argue that Intel should be able to gain. However, AMD is not actually that far behind. The latest reviews that I've seen suggest that K8 is about 10% slower per clock than Intel. However, since K8's are priced lower per clock than Intel this makes the AMD processors about the same except for FX-62 which is more expensive.

That AMD's processors are priced about the same as Intel's for the same power has again caused a lot of consternation from Intel fans. Some have even insisted that AMD should price its X2's below the lowest priced C2D, the E4200 but this has not happened. I still see the arguments popping up, groundless though they may be. It seems that some just cannot understand that the fact that E6700 is faster than FX-62 doesn't mean that all C2D's are faster than all X2's. The AMD X2's priced similarly to E6300, E6400, and E6600 have similar peformance for most tasks. And, there is the neverending misuse of benchmarks to prop up smoke and mirrors arguments about C2D. For example, it is understandable that video conversion benchmarks keep all of the data in memory to test the processor. However, this does not mean that users will see the same speed when the data is pulled off the harddrive under actual use. This is true of common user applications which spend most of their time waiting on the user. There are similar problems with tests that don't use all of the cores, benchmarks that show big gains because of cache, and benchmarks that are based mostly on the graphics processor. For example, a benchmark might be very speedy due to cache when it is the only thing running but this doesn't reflect a real computing environment at all where that same process would have to make do which much less cache.

Perhaps because of the lack of gains by Intel there has been a flurry of negative rumors about AMD lately. This includes ideas that AMD is behind in 65nm, is having problems with 65nm, and won't deliver K8L on the desktop until Q4 07. AMD has said that 65nm production is at 25% at the end of 2006. It is impossible that AMD could be at this level without having shipped something like 8 million processors. Some of these would have to have been delivered early enough to be out before year's end yet rumors persist that AMD has delivered no 65nm chips. The second idea comes from the fact that the 65nm chips are not half the size of the 90nm chips. This has led to some rather wild speculation that AMD has added new undocumented sections to the die or that AMD is having problems with 65nm. Actually, neither of these are true. Although 65nm is capable of producing die features that are half the area of 90nm you can't simply shrink 90nm transistors like you would a photograph. If you just shrink the 90nm transistors then at some size they will stop working properly. The first production with 65nm is only a partial shrink because this is the first round of 65nm transistors. These will get tweaked better for the 65nm process in revision H (K8L) and will be able to be reduced more. The 65nm dies that AMD produces at the end of 2007 should be shrunk more than what is being produced now. Again, this is not unusual as AMD did the same thing with 90nm.

As to the idea that AMD won't deliver K8L on the desktop until Q4 07, AMD has contracts with one supercomputer that will be running by end of Q2 07 so K8L production has to start in Q2. Desktop versions of K8L will come out in Q3. Apparently, the fact that the intitial supply of K8L is tight because of preordering has led some to believe that production won't start until Q3. Then apparently the logical assumption is to push back the desktop production to Q4. People have likewise pushed Intel's production of 45nm up from Q4 to Q3. This would then have 45nm being released from Intel before AMD has any desktop K8L's. However, not even Intel is saying this. Real production of 45nm by Intel in Q3 07 is unlikely.

And, what would rumors be without speculation of a point to point bus for Intel? Supposedly, Intel will release a point to point bus (CSI) in Q3 07. Actually there is no indication of this at all. The earliest that it looks like Intel could release CSI is early 2009. It looks like people are getting CSI confused with Intel's quad FSB chipset which will finally allow Woodcrest to be 4-way. The confusion seems to come because Intel insists on calling this a high speed bus rather than a Front Side Bus. Other than the name change it looks and works exactly the same as the current Intel FSB's that connect the processors to the northbridge. A true use of CSI would require an onboard memory controller like AMD has plus a new socket since you cannot run a memory controller on the current socket 775/771. There are other cards to fall yet. For example, AMD is releasing a faster Athlon 64 which should match E6700. This could be offset by a faster 3.2Ghz Conroe but only if Intel slides its prices down one notch. I'm sure Intel will do this eventually but perhaps not in Q1 07. There have been rumors that Intel will have 3.5Ghz chips by end of 2007 but then again the rumors insisted vehemently that Intel was going to have a 3.2Ghz Conroe at launch and this hasn't happened yet.

So, 2007 becomes a race. Intel will ramp C2D up from 25% while AMD continues to ramp 65nm on FAB36. Then once the new bump and test facility is completed AMD will begin taking FAB 30 down for conversion to 45nm. The best situation for Intel would be that the market expands rapidly which would allow Intel to grow faster than AMD since AMD will be capacity limited through most of 2007. With normal expansion, Intel is likely to continue to slowly lose share to AMD. This may seem unfair however AMD has cultivated better relationships with customers than Intel has in the past. By the time Intel has demonstrated any change AMD will have K8L. Further, Intel's gains have mostly been on the desktop which is the least profitable area. AMD has gained in mobile which is more profitable, has produced fewer low end Semprons, and is still holding onto the top 4-way server market. Intel won't be able to move into 4-way until Q3 07 at the earliest when it finally has a chipset with 4 FSB's. Further gains by Intel on the desktop will most likely be offset by gains from AMD in both mobile and corporate areas. I would expect AMD to end 2007 with a gain of about 2.4% in volume share.

The problem for Intel today is that this isn't like 2002 when Intel was able to grab back large amounts of share. Intel now has to deal with an AMD that has a credible name in servers, a growing presence in corporate, total system solutions from the ATI purchase, a growing presence in mobile, and both reduced costs and increased capacity with a second 300mm FAB. The situation would be completely different today if AMD still had only FAB 30 and if this were the case all my money would on Intel. However, with FAB 36 coming up in capacity and AMD's uncanny ability to do rennovation without taking a FAB completely offlline I just can't see anyway for Intel to take advantage. If the market expands quickly then AMD would gain less volume share but revenue would be better for both AMD and Intel. I'm still expecting Intel to have to do more divestment and probably another reorganization in 2008. Overall, however, 2007 should see Intel's and AMD's product lines and graphics support move much closer. By end of 2007 Intel should have something to compete in 4-way but it will have increased competition in 2-way and the desktop. 2008 will then bring new mobile competition from AMD as well as increased server and graphic competition. Intel will have to be on its toes in 2008 to stay competitive and 45nm by itself won't be enough. Looking further ahead Intel proponents are quick to point to CSI but Intel will then be facing an AMD with two functional 300mm FABs, tightly integrated graphics, and a modular core for easier upgrades so 2009 will still be highly competitive. Worse still for Intel, with 30% of the cpu market for AMD, Intel's monopoly will be broken. So, appearances in 2007 should be better for AMD and Intel should be able to climb back up to the 2005 processor revenues but mainly the advances by both Intel and AMD will be good for consumers.

Sunday, December 24, 2006

Music -- The Collapse Of An Industry

I've seen several Time/Life commercials for music collections lately. The problem is that while these collections don't really solve a problem they do tend to point out just how awful the quality of music has become. Somehow, music producers have slowly disconnected from their own customers. Today, the quality of music is so bad that there is no strong incentive to buy it. However, at the same time the channels for delivering classic music are so poor that this is cost prohibitive. Today, there exists a huge demand for music that is not being met while music producers are either unaware or unable to meet the demand. According to the recording industry's own data, music sales are down 33% by volume.

I've recently seen commercials for Time/Life's 70's collection. This is 10 CD's with 150 songs for $150. The problem is that this pitiful collection doesn't even scratch the surface of 70's music. The minimum collection would be 200 songs from each year which would amount to 2,000 songs and presumably 134 CD's for $2,000. This is not an offer that is likely to appear from Time/Life. I doubt though that anyone at TL even realizes or cares how insulting this notion is; 150 songs just does not and cannot do justice to a decade of music. 200 per year would be a bare minimum. A more reasonable number would be 10 new songs per week and about 500 songs total per year. This would give us 5,000 songs total for the 70's and would take over 300 CD's. This just isn't practical. How many people are prepared to fork over $5,000 at $1 per song to have a genuine collection of 70's music? And, what about the 60's and the 80's? How many people have a spare $15,000 for music? And, my estimate of 10 songs per week is just a ballpark figure; we could still be scraping off some very good music with this number. Perhaps 10,000 would be better but this of course would just make the disparity even greater. And, TL is not unusual on its pricing. Assuming that a 10,000 song 70's collection would even be available at iTunes the price would basically be the same at 99 cents per song. It is clear that cost is a prohibitive problem with delivering classic music.

I can identify with music from the 40's, 50's, 60's, 70's, and 80's however, it appears that there is a definite deterioration starting in the 90's. This trend appears to have bottomed out and has now stayed at the same low level since 2000. To put it simply, the quality of music today is only a fraction of what it used to be. I wondered what could have happened to throw the music industry so badly off the track that it had followed for half a century. It appears that the problem is radio. Although music producers create and distribute the music it has been up to radio stations to make buyers aware of new music. This was never a problem in the past as the great number of independent radio stations were all too eager to cater to many different tastes in music. However, during the 90's these independent radio stations were bought up by media conglomerates and the process began to short-circuit. Rather than having independent programming, the play schedules became homogeneous across dozens of stations from city to city. As consolidation continued the number of choices became less and less. This had a chilling effect for music producers since there were fewer and fewer independent radio stations to promote new music. In the past, a good song would always be picked up and played by a disc jockey somewhere and there are many examples of hit songs that started from just one station. With independents gone it became important to music producers to have music that was acceptable and this has led to drastically reduced risk taking by producers both in terms of style and in terms of costs.

The rise of rap music was not so much due to embracing a new genre by music fans but rather a genre that has been pushed by producers due to cost considerations. With typical music production studio time is very expensive and requires long hours laying down multiple layers of instrument and vocal tracks and then many more hours mixing and adjusting the sound until it's right. However, rap and hip hop typically use pre-recorded music so this removes the studio time spent recording and mixing the instrument tracks. Rap and hip hop also typically do not have as many vocal layers. For example, a traditional song might require four music layers for percussion, rhythm, lead, and solo instruments using half a dozen musicians. If instruments like saxophone or trumpets are used then it will take more layers because these instruments tend to have different volume levels than other instruments. These layers can be much more elaborate as when The Eagles used symphony music for Wasted Time and Fleetwood Mac used a marching band for Tusk. The vocal tracks can also run to many layers since there will be verse, chorus, and backup vocals and sometimes multiple layers of the same singers for self harmonies as was common with The Carpenters, Air Supply, and Janet Jackson. One album can involve several song writers, many musicians and vocalists. In contrast, A rap album often involves pre-recorded music with one song writer and one vocalist. Even when these are stretched it doesn't usually involve more than a drum or rhythm loop and one or two other vocalists. With much less effort involved, the cost of producing a rap or hip hop album is only a fraction of the cost of a regular album. This has made rap and hip hop much more attractive to music producers since the up front costs and therefore the risks are much less.

It isn't difficult to tell that radio stations in general have declined. When I was in high school there were two local radio stations that played popular music. These are both gone. Popular music was also available on AM however the AM band has almost entirely been replaced by talk radio. Judging from what I have heard on country music stations it is possible that country music has fared better. It is clear, however, that both variety and quality of popular has drastically declined. I am just not seeing bands today of the caliber of The Eagles, Fleetwood Mac, Journey, ELO, Foreigner, Chicago, Steely Dan, Moody Blues, Temptations, The Doors, Spinners, Beetles, BeeGees, Air Supply, Reo Speed Wagon, Styx, Duran Duran, Jefferson Airplane, Herman's Hermits, Queen, Hall and Oates, Heart, England Dan and John Ford Coley, Credence Clearwater Revival, Crosby Stills and Nash, Police, etc. This list is just a tiny sampling of very good groups and their replacements are nowhere to be seen.

The need to contain costs has led to fewer bands in favor of solo artists as well as the rise of rap and hip hop. Likewise the strong need to reduce risks has greatly curtailed variety in music that is produced today. This drastic reduction in both quality and variety has caused a sharp drop in demand for music which has in turn created a downward spiral for music producers. With less demand, there is less chance of having an investment payoff and this in turn leads to less risk taking which leads to less variety and quality and ultimately even less demand. The demand is actually there and as strong as it has ever been; it's just that the demand is not for what is being offered. The present situation is incredibly strange as there are both a vast number of talented songwriters, musicians, and singers who want to fill the demand along a huge pool of consumers who badly want good music. Supply and demand have become so out of balance that it is a bit like New Orlean's being below sea level. The only thing currently holding back supply from demand is the unnatural damning effect of the radio conglomerates and the ever more conservative music producers.

The traditional mechanisms of radio involves commercials and advertising support. However, with less variety and quality there is less reason to listen to a given radio station. Consider that while the radio conglomerates are trying to scrape by with less variety I can listen to 70's music with no commercials on my cable all day long for only about $10 per month extra. This is drastically less than it would cost to purchase the music outright from iTunes or to buy a TL collection. The music industry seems to be stuck with the idea of total ownership rather than temporary use of a song. When I listen to music on cable it's only temporary; I don't keep or own the music. I'm not sure if it is the idea of piracy that prevents the music industry from a more reasonable temporary pricing rather than outright pricing. With 1,000 songs per year and 4 MegaBytes per song this would require about 120 GigaBytes of storage to have a good collection of 60's, 70's, and 80's music. This is certainly within the range of common harddrives today so storage would not be a limiting factor. However, organization would be. Having the artist, song title, albumn title, and year of release for 30,000 songs would be no minor task. Without some type of database to keep track of all this information it could be nearly impossible to find the songs that you want to listen to even with your harddrive stuffed full of mp3's.

However, if the music industry were actually doing its job then this wouldn't be a problem. The music industry could easily maintain records of music titles, artists and groups, year of release, along with much more more information such as lyrics, band members, songwriters, covers, etc. It should be possible to find a band by knowing the name of one band member or to find a song by knowing some of the lyrics and it should be quite easy for individual playlists to be compiled for each listener. Perhaps this will happen at some point. Perhaps we will have distribution via some type of WiFi, and music will bypass broadcast radio altogether. This could allow more tailored advertising and sampling of new music based on your established tastes. I'm thinking that something like this will have to happen because I cannot see this vast gulf between talent and demand continuing forever nor can I imagine that music producers will be dumb enough to leave themselves cut off from their source of income indefinitely. However, until something like this happens presumably music production will stay at the low level of quality and variety that it is today and radio conglomerates will continue to choke off supply.

Friday, December 15, 2006

Intel Versus AMD: The Truth Unfolds

It has not been easy to find out the real story of how Intel hardware compares with AMD hardware. The reviews from places like Anandtech and Toms Hardware Guide have been so sloppy and unprofessional that no useful information could be obtained. However, after a second review at Tech Report there is finally enough accumulated data to draw some conclusions.

Often when I mention the horrible state of hardware testing today I get some flack from Intel fans who really want to believe that any published test results that show Intel favorably must be true. I suppose there have been a number of people who have suggested that I have been somehow denying reality by not embracing these results. Of course, these people don't understand at all how I think. Poor testing doesn't mean that AMD is better; it just means that you don't know. With proper testing, the results could show that Intel is even better than what was shown with poor testing. I suppose since I'm now talking about testing done at Tech Report some must be wondering if I am endorsing Tech Report as a good fact source. Well, not exactly. The problem is that it has taken two reviews (and sometimes a third) to get the facts at Tech Report. If the testing had been truly prossional then we would have been able to get what we needed from the just one review. But, this is not the case; it is necessary to look at two or more reviews to find out how reliable the tests themselves are. What this means is that some of the data published by Tech Report does not fully show what is happening with the hardware and therefore this data can be misleading (and the author's comments can be incorrect as well). However, by using more than one set of tests there is finally enough information to draw some conclusions. I suppose this is at least better than most other review sites where the testing still shows essentially nothing even with two reviews in hand.

One of the issues that keeps getting tossed out is the idea that Intel processors now draw a lot less power than AMD's. This seems to be due in part to a desire to throw off the very bad reputation that the Prescott cored processors had. Yet, there is no indication of any big advantage for Intel. The highest clocked chips from both companies are rated at 120 watts and the second level from both companies runs about 80 watts. You have to add a bit to the Intel numbers because their TDP isn't a maximum like AMD's and you also have to add the power draw for the Northbridge which is included with the AMD numbers. However, Intel's C2D processors have much better power management than AMD's current Revision F and 65nm Revision G processors so this tends to even out. The sole advantage that Intel currently has for any class is for their quad cored processors and this is only because AMD doesn't have a quad core out yet. You could however argue that if you rate processors by SSE performance per watt then Intel is better. This is true if you use SSE most of the time such as for heavy duty rendering or scientific or engineering calculations. But, for most areas Intel and AMD are about equal. This shouldn't change until late 2007 when Intel releases 45nm chips.

The first benchmark is Cinebench which renders a graphic scene. Surprisingly, Intel and AMD are almost dead even on this benchmark. This is not due to AMD's greater memory bandwidth because this test is not memory intensive. I would have expected Intel to be faster but this is not the case. At the same clock speed the two are tied. This benchmark also does not scale very well; perhaps the code will be better by the time AMD has its own quads out.

The next is Pov-Ray and once again the results are very close; Intel trails by about 4% at the same clock. Either Pov-Ray does not make use of SSE or AMD is gaining more from the 64 bit code. This application is not memory intensive but it does show good scaling even with 8 threads. A dual socket Clovertown Xeon 5355 system should deliver about the same performance as a quad Opteron 8218 system and a single socket Kentsfield QX6700 should be about the same as a dual socket Opteron 2218 system.

With 3DS Max scanline rendering there are two factors. First of all, for a comparable system, Intel is 25% faster at the same clock than AMD. This is probably due to better SSE. Secondly, this code does not scale very well. Intel shows a 22% drop-off when scaling from 2 threads to 4 while AMD shows a 29% drop-off. This looks suspiciously like poor NUMA support. Finally, in contrast to the scanline rendering, 3DS Max's ray tracing code is so poor for multi-threading that it isn't useful at all for comparison beyond 1 core.

In the Quad FX review the Valve Source engine particle simulation numbers are all over the place. I've tried to see if anything can be gleaned from this information but unfortunately there are no matching numbers in the Clovertown review. This is a good example of a benchmark that less knowledgable fans might point to as an Intel win but these numbers are far too erratic to tell anything. The Valve VRAD map build however shows Intel 26% faster at the same clock. There really isn't anyway to tell if this is due to cache, integer IPC, or SSE but Intel is clearly faster. This will be a good benchmark to repeat with K8L. Since K8L should increase cache, integer IPC, and SSE the results should be much closer.

3DMark is not in the Clovertown review but after comparing with an older review with C2D E6300 it is clear that 3DMark 06 CPU Test 1 is good and not simply gaining from larger cache; Intel is about 10% faster.

The MyriMatch test is interesting. It is clear that this code is terrible with multi-threading and quickly loses efficiency as threads increase. However, the numbers drop off not only with more threads but with faster clock speeds as well. Worse still, the Intel numbers show variances of 5% when increasing clock by 15% and the AMD variances are even greater at about 8%. This leads to a very unreliable spread showing Intel is somewhere between 3% and 27% faster. Since I did not see this variance in the other tests and there would be no reason for Tech Reports to have reconfigured the systems during this test I would have to say that the code is at fault and the test is therefore useless.

The Euler 3D computational fluid dynamics test is a good example of how misleading the data can be and how equally far off the author's interpretation can be. He claims that the code is both floating point and memory intensive. Yet, when comparing the two sets, there is no evidence of any drop-off due to memory. With double the memory bandwidth, the dual socket woodcrest system is actually slower because of the higher latency FBDIMM's. There is no gain at all from the higher memory bandwidth, therefore the code is not memory intensive in the sense of saturating the memory bandwidth. Intel seems to be about 18% faster. However, some of the scores are far worse than 18% and this again brings up suspicions of NUMA problems. Looking more closely, we see that in spite of claims that the Quad FX tests were done using a NUMA compatible OS, there is clear proof of the opposite. Quad FX and the dual socket Opteron 2218 system are very similar as both are dual core on dual socket. And, their scores are almost identical even though the FX is clocked 15% faster and uses faster memory. Apparently, the interleaving that was used with the Opteron is making up for the faster speed of FX, but this could only happen if the FX were running without NUMA support. This means that for any code in the FX review that uses large working sets the FX processor will show an anomalous dropoff. This is why one review was not enough; you can't see this without comparing with the second review. So, Intel is 18% faster rather than the 73% as would be suggested by the FX numbers. Even with the interleaving hack this very poor number would have been cut in half and then half again with proper NUMA support.

The Folding @ Home series is interesting. But, there are errors in the totals for both reviews. The actual totals are pretty close and would give Intel about 6% more speed. However, what we really want is the best rather than totals or averages and this would give Intel about 25% more speed. However, it is stated that the same code for AMD uses 3DNow! rather than SSE so the actual difference could be less. However, I think 25% due to better SSE is reasonable. This is a benchmark to watch because it should have an increase with K8L unless the code is not actually making use of the AMD hardware.

The Sandra benchmarks are both good and bad. The Multimedia Integer tests are way off. I'm not sure what they are testing but whatever it is does not show anything real about the processors. If these scores were taken at face value Intel would be 253% faster in integer. Or to put this a different way 1 C2D core would be 89% as fast as 4 AMD cores. This is yet another published score that unknowledgable fans might tend to mention as a sign of the phenomenal speed of C2D. However, this is obviously impossible theoretically, architecturally, and from the fact that no other test shows speed like this. Therefore, we have to ignore this test. However, the Multimedia Floating Point test looks good. Intel is 79% faster. This is reasonable since Intel can easily do 100% faster with SSE and could theoretically, with perfectly matched data, actually hit 4X AMD's speed. However, this is not a likely thing to occur. This is another benchmark that should become about equal with K8L. Perhaps by the time AMD releases K8L we will know what is wrong with the Integer test.

The Windows Media Encoder shows a significant drop with 4 threads, but it drops about the same for Intel and AMD. With the original code Intel had a 23% lead, however with the newer 64 bit code this lead drops to only 15%. This is true for both 2 and 4 threads.

PicColor contains some anomalous scores which lessons confidence in the set's overall accuracy. The extremely low cpu usage graphs add to the uncertainty. The best I can say is that for most tasks Intel seems to be about 12% faster but is about 60% faster on floating point.

Panorama Factory is not memory intensive. It shows poor scaling with 4 threads and truly horrible scaling with 8 threads. However, the scores appear consistent. Intel is 7% faster. The only odd score is the FX-74 score which is slower (12% instead of 7%). This seems to be due to NUMA problems for the large datasets. Apparently, interleaving is enough for this task but FX-74 without even that shows a drop-off.

Tech Report has used three different methods of power consumption testing with the three reviews I referenced. This somewhat diminishes the value of the power consumption tests. For example, the Quad FX review does show some advantage for Intel, however the load application is Cinebench which has poor scaling. In the Clovertown review the load application is Pov-Ray which scales well and the power consumption is very similar. Realistically, Intel only seems to gain with lower power draw with its Kentsfield QX6700 processor. There is no real difference with dual core/single socket between Intel and AMD nor with dual socket.

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. For example, I usually match X2 4200+ with E6300, 4600+ with E6400, and 5000+ with E6600. Yet, it seems for many tasks, Intel only appears to be 10% faster. This would mean that 3800+ would match with E6300, 4200+ with E6400, and 4600+ with E6600. This would make 5000+ equal to E6700 but AMD would still have no procesor to match X6800. This seems to be about right if we exclude SSE. However, if SSE is needed then Intel is the current bargain.

Upgraded To Beta Version Of Blogger

I recently discovered that my Blogger ID had stopped working and I was unable to log into other blogs. My blog ID still worked on my blog but not others. I then discovered that I could log into other blogs with my Google ID so this gave me a good idea what the problem was. Apparently at some point all of the old blogs will be changed to the newer Beta version. The Beta version uses a Google ID rather than a Blogger ID. I already had a Google ID since I had to create one for AdSense but I had never actually upgraded my blog. The upgrade messages showed up everytime I first logged in.

When my Blogger ID stopped working it became clear to me that Google has begun changing the login format and that this was why my old Blogger ID was not recognized on other Blogs. I have no way of knowing if this effects all Blogger ID's or just people who have blogs of their own. For me, it was necessary to upgrade my blog to the Beta version to straighten out my login ID. It seems to work fine now just with the Google ID instead of the original Blogger ID. I have to say though that the upgrade process is confusing because you are faced with a screen that asks if you have a Google account already. The choices are Yes and No. The problem is that there is no button on the Yes side so you can only click No. I didn't really want to click No since I already had a Google account but I had no other option. If you click the No button it will take you to another screen where you can then either create a Google account or login to your existing account. So, the option of using your existing Google account is there on the second screen but Google could have made this so much clearer.

Now I am wondering with my blog upgraded to Beta whether other people might be having trouble logging in. My suggestion would be that if you don't have a blog and your Blogger login doesn't work that you try creating a Google account. If you do have a blog then you might have to upgrade to Beta like I did. You can get some information about this from Blogger Help. Thank you.

BTW, I need to leave this article up for a little while to let people know what is going on. However, I'll mention that I'm working on my next article which will probably be called,

Intel Versus AMD: The Truth Unfolds


Tuesday, December 12, 2006

Walking Into A Mirror

The most significant development this year in the competition between Intel and AMD seems to be that the days of moving in different directions are over. K6 was not so different from PIII. However, when AMD moved to K7 and Intel moved to P4, the architectures diverged dramatically. As AMD continued with K8 and Intel moved to Prescott, the architectures became even more different. By 2005, in terms of both architecture and corporate structure, there was little overlap between Intel and AMD. This year, with the introduction of Core 2 Duo Intel has pulled its architecture back in the direction of K8. Likewise, AMD is moving in Intel's direction. What we will see through 2007 and 2008 will be the convergence of the two companies both architecturally and structurally.

In 2005, AMD and Intel were as different as they had ever been since AMD started making X86 processors. AMD had only a single 200mm FAB compared to Intel's many 300mm FABs. And, having divested of Spansion, AMD was strictly processors with a few server chipsets in contrast to Intel's diverse interests including its large chipset and embedded graphics business. In terms of cpu architecture, Intel had a long pipeline, virtual HyperThreading, high clock with low IPC design. AMD's short pipeline, dual core, low clock with high IPC design was almost completely the opposite. Intel's MCM Smithfield procesors were a small step in AMD's direction. However, in 2006 with C2D Intel moved to a short pipeline, high IPC, true dual core design and dropped HyperThreading. The architectures are still different but not nearly as different as they were last year and they are still converging.

Just as Intel moved closer to AMD by shortening the pipeline and increasing IPC while extending the architecture to 64 bits so too AMD nows moves in Intel's direction. Core 2 Duo has a dedicated stack unit and AMD moves closer to this with its sideband stack optimizer. AMD improves instruction reordering and out of order loads moving it closer to Intel. AMD now has data dependent latency on the divide instruction as Intel had already done. AMD stretches SSE width to a full 128 bits doubling the speed of SSE operations. It doubles the bandwidth of Prefetch and the L1 bus. Again, these are changes that Intel had made from Yonah (Core Duo) to C2D. AMD might have a small advantage here due to twin L1 buses rather than just a wider L1 bus.

Intel's architecture has been better with DDR2 because of the larger cache. AMD is now moving to a split memory controller where each half can act independently which is what Intel has now with two memory channels. AMD also increases cache size by adding shared L3.This along with improved memory scheduling, a better write burst mode, and out of order loads more than makes up for the higher latency of DDR2. We should see a big improvement for AMD with DDR2 and into DDR3. This again, moves AMD up to where Intel is now.

AMD also moves in Intel's direction by adding some inclusive functionality to its cache while also adding shared cache as Intel has now with C2D. AMD is expanding the number of FastPath instructions. This is important because the decoders can do 3 FastPath instructions or 1 complex instruction per cycle. The more instructions that are fastpath the better. This again makes up for some of the improvement that Intel gained by adding a fourth decoder and macro-ops fusion.

Overall, AMD moves much closer to Intel's C2D with the changes in K8L. AMD will add to its already greater memory access speed and match or beat C2D in terms of SSE. However, the changes don't quite add up to C2D's 4 instruction issue and Macro-ops fusion. I would guess that AMD will remove about 3/4 of the current IPC gap. This would mean that a K8L of 3.1 Ghz would match a C2D of 3.0 Ghz. This is much better than today where it would take a 3.5 Ghz K8 to match the same 3.0 Ghz C2D.

Both AMD and Intel use independent L1. AMD's independent L2 plus shared L3 should be pretty close to Intel's large shared L2. The main differences today are that Intel's C2D still does not allow I/O access above 32 bits and I'm not certain that C2D supports addressing beyond 36 bits. AMD has the advantage with an onboard memory controller and point to point interface. It looks though like Intel will defintely move in this direction with its own onboard memory controller and CSI point to point interface on a new socket in early 2009. AMD has also given indications that it is trying to shorten Intel's smaller process lead.

Structurally, AMD has taken big steps in Intel's direction by creating a 300mm FAB and purchasing ATI. ATI gives AMD a directly competitive position with Intel in terms of chipsets and embedded graphics. AMD will now be able to satisfy the needs of customers who want a complete solution from one manufacture as Intel has provided for some time. AMD will now compete with Intel directly in both embedded graphics for notebooks and embedded graphics for corporate customers. At the same time, Intel has begun divesting of underperforming sections of the company and trimming down toward the main cpu business closer to where AMD already is. AMD's single 300mm FAB greatly boosts its production capacity while reducing cost. Upgrading FAB 30 to 300mm will continue to move AMD in Intel's direction.

There are probably AMD fans who are hoping that K8L will allow AMD to leap ahead of Intel's C2D in 2007. Similarly, I imagine there are Intel fans who are hoping that Intel's move to 45nm will allow Intel to stay ahead of AMD. However, neither of these scenarios seems likely. K8L should be pretty even with C2D architecturally rather than a leap ahead. For process technology the rumors abound with stories of 4.0Ghz clock speeds for Intel while decreasing power draw. These same rumors often limit AMD to as little as 2.9 Ghz putting AMD one full Ghz behind Intel. This would put Intel significantly ahead of AMD if true but it doesn't seem likely. AMD shares process technology with IBM. It is certainly in IBM's interest to be competitive in terms of process. So, it seems likely that if something were lacking that would allow Intel to gain significant advantage IBM would find another company with additional process technology that could be licensed. I'm reminded of the many rumors that Intel would have 3.2Ghz C2D chips at launch but these are still unreleased. I'm also reminded that after Revision F of K8 was released there were claims that it was actually slower than Revision E. It seems that optimism tends to follow Intel while pessimism follows AMD regardless of what each company actually does.

I expect AMD to continue to slowly gain share at the rate of about 0.66% per quarter. I expect AMD and Intel to be roughly equal on the desktop in Q3 07. Likely Intel will have to address AMD's new dual socket FX once AMD has quad core chips for these. If Intel did not address this area they would have nothing competitive. It also looks like Intel will finally have a 4-way server C2D offering in the same timeframe. However, just as AMD's quad FX won't reach its full potential until AMD has quad core FX's, Intel's 4-way server chipset won't really pull Intel even with AMD. Clearly, this will be a stopgap until Intel can release a new socket with integrated memory controller in 2009. However, Intel has no choice but to offer a stopgap chipset since the P4 Xeon will be completely outclassed by K8L. Intel's processor will be fine; it will be the chipset that is stopgap. Intel's quad FSB northbridge will be slower and more expensive than the equivalent Opteron 4-way but it should be close enough to stop Intel's 4-way server share loss. Intel will probably get a nice boost with 45nm technology that will give it an edge on the desktop late in 2007 but AMD will follow with its own 45nm technology by mid 2008. However, any advantage that Intel might have will tend to be countered by AMD's increased competition with embedded graphics. But this won't be the only change. Increased revenue gives AMD more money for R&D and AMD's advantage with APM will come into its own once AMD has two 300mm FABs. My guess is that Intel will also be taking steps to try to match APM and will probably have to do a second reorganization in 2008. AMD for its part may begin construction of a third FAB, possibly after mid 2007.

We know that Intel will move to an Integrated Memory Controller moving it towards AMD. It would seem reasonable that AMD must be looking into Intel's 4 instruction issue and macro-ops fusion to gain future speed. We also know that Intel will eventually use SOI as AMD has been doing for years now. By 2009 AMD and Intel should look much more similar than they did in 2005. I imagine the two will continue to converge as well go into 2010. For example, Intel is moving to a 2 year core design cycle while AMD is moving to a modular core. Intel's approach is less efficient than AMD's however Intel has more engineers to throw at the problem so I would imagine the results would roughly match. If 4 instruction issue really is a good idea then I would expect AMD to add this perhaps in late 2009. There will be a slight difference in process with Intel using high K dielectrics. I don't see any problem with this. Intel only needs to find a high K material with a suitable bandgap for this to work. The only area that I'm not so sure about is Intel's tri-Gate design. From what each company has said it looks like the tri-Gate will not be able to be shrunk while AMD's finFet will be. Still, one would assume that if the tri-Gate can't be shrunk that Intel would follow AMD and use a 3D dual gate like AMD's finFet. If AMD does build a third FAB the X86 processor market should turn into a true duopoly with much more stable prices and production. The big question is whether the market will reach duopoly status with just two 300mm FABs at AMD. However, it seems a certainty that into the foreseeable future AMD and Intel will continue to converge.

Sunday, December 10, 2006

Misinformation Rolls Onward

I had hoped that hardware review sites like Toms Hardware Guide and Anandtech would somehow rediscover the integrity they used to have. I've hoped that financial analysts would stop measuring Intel with a shorter yardstick than they use for AMD. However, even a brief look at recent articles shows that bias towards Intel is very much alive and well.

In this THG Clovertown article we see several problems right away. For example the only two compilers used are: Intel C++ Compiler Version 9.1.030 and Intel Fortran Compiler Version 9.1.029. This guarantees that Intel will have an advantage because the Intel compiler purposefully skips optimizations on non-Intel processors. Secondly, we see that the bulk of the "tests" are either single or dual thread. I'm still baffled why anyone with common sense would test a quad core processor with single threaded code. This is a bit like testing a 500mph test vehicle in rush hour traffic in L.A. rather than someplace more suitable like the Bonneville Salt Flats. I suppose you could make the same lame argument that L.A. traffic is more like real driving conditions but a test is useless if it is artificially limited as so many of the recent tests have been. We also see that in the supposed "load" test that the testers fail to actually achieve any load on the processor. Intel gives the TDP of the 5355 as 120W while the 5160 is only 80W. Yet in the THG chart on page 11 the 5355 actually draws less power. I had strong hopes that THG might get back on track during 2007 but this level of incompetence or outright bias is hard to break. THG remains useless for real testing and benchmarking.

Another interesting article was this Economist piece. The theme of the author is clear as the article is entitled The Empire Strikes Back. Backing this up is a supposed graph showing that Intel's market share is on the rise and growing faster than AMD's. The only problem is that this graph is bogus. To get this imaginary increase they include numbers from VIA whose cpu's sell at a price below both Sempron and Celeron and therefore have no effect on Intel and AMD processor sales. The numbers for VIA were temporary inflated because VIA was dumping large numbers of processors that were End Of Life. By adding these anomalous VIA numbers in they make Intel's Q2 share look worse and give the impression of a changing trend. In reality, Intel's share has been slowly dropping while AMD's has been slowly rising. The change is roughly 0.66% per quarter. This trend should continue and if it does AMD will indeed arrive at 30% volume share by the end of 2008. I guess the problem is that if Intel has been slowly losing share for the last six months it would be impossible to use those snappy catch phrases in the article like, "right-hand turn", "turning the wheel", "successfully navigated the transition", and "Intel cemented its comeback". I guess the true phrases like, "Intel continues slow loss of share" or "Intel's reorganization was not deep enough" just wouldn't generate the same kind of excitement. In reality there has been no big turnaround in share for Intel in spite of the introduction of Core 2 Duo. Today, Intel continues to slowly lose share to AMD. Perhaps it is time to give up on the financial analysts as well.

Over at Overclockers Ed is claiming that AMD will only hit 2.9 GHz with dual core on 65nm. I'm not really sure where Ed got this information from. Common sense would suggest that if AMD can hit 3.0Ghz with FX-70 on 90nm then they should be able to do at least a little better with 65nm. If Ed had speculated 3.2Ghz I think that would be reasonable. Even 3.1Ghz would be a bump over 90nm speeds. However, I think speculating a speed that is lower than the speed for 90nm is just absurd. Yet in the same article he speculates that Intel will hit 4.0 Ghz with dual core on 45nm. Considering that Intel has not yet shown or announced anything faster than 3.0Ghz I would have to say that this is optimistic. The question is why the optimism always seems to be in Intel's favor. Overlockers is another website that is unreliable for a balanced look at Intel and AMD.

A website that is usually much better, Ars Technica, however showed some serious errors in their Barcelona article. This is the faulty paragraph:

It does not seem that K8L will catch up to Conroe in terms of the theoretical peak number of 128-bit SSE operations per cycle, however. K8L's two floating-point/SSE pipes give it two 128-bit SSE ops/cycle, and its FSTORE pipe can do another 128-bit SSE move per cycle, for a total of three per cycle peak. This is half of Conroe's peak theoretical throughput of six 128-bit SSE ops/cycle.

Conroe cannot do 6 SSE ops per cycle. It is true that sometimes Conroe can do a simultaneous SSE multiply/add in a single cycle however 6 ops is pure fantasy. Where the 6 ops number comes from would be two simultaneous multiply/adds plus one load plus one store. However, the actual datapath to L1 cache can only handle two 128 bit words at a time. So, a more realistic number is two 128 bit SSE operations per cycle which interestingly is what K8L will be able to do. It is true that sometimes Conroe could go faster by executing combined multiply/add instructions but the article fails to count the L1 datapath advantage that AMD has over Conroe. My best guess is that AMD will actually come out slightly ahead of Conroe on SSE.

I think that there are many who are still hoping for Intel to pull decisively into the lead again as they did in 2002. However, this is not happening and won't happen. The true situation is that AMD and Intel are becoming more and more alike both in terms of corporate structure and in terms of processor design. I'll talk more about this in my next article, Walking Into A Mirror