   |
|||
|
|||
|
:
News
: Archives
|
|
|
|
|
|
|||
|
|
|||
|
|
|||
|
July 20, 2007 Yesterday AMD posted revenues of $1.378 billion, and a loss of around $600 million. While the loss is larger than expected, revenues beat the street by a pretty hefty margin. Speaking of margins, AMD has stated that they have increased margins from last quarter's 31% to about 34%. The news of revenue growth and increased margins has had a positive effect on AMD's share price in after hours trading. In pre-market trading the stock is up $0.34 from yesterday's $15.78. Obviously this past quarter was a bittersweet one for AMD. It appears as though they gained back some marketshare from Q1 levels, but their ASPs have gone down (as evidenced by the low, low price of Athlon X2 products and the lack of a high end product over $200 on the desktop). AMD has also lost a significant amount of Intel chipset business, and their latest Intel supporting chipset the RD600 is only being carried by one manufacturer (DFI). When AMD bought ATI, Intel switched to SiS to support the low end integrated market of Intel branded motherboards. On the bright side Athlon 64 supporting chipsets have increased, most likely due to the very impressive 690G integrated product. So far most major manufacturers have released a 690G product, and this chipset helps to push AMD CPUs out the door as well. Considering it is fully Vista Aero capable, and performs at the top of its class with excellent video playback features, this chipset is becoming very popular in the desktop market. The M690 has achieved quite a few design wins as well in the mobile category, and when combined with the low power Turion processor, achieves very impressive results in terms of power consumption, battery life, and 3D and video performance. The only real downside to this chipset is that it is based on the X700 series of graphics processors, which supports up to DX9 SM 2.0. NVIDIA's integrated parts support SM 3.0, as does Intel's X3000 series (with the latest beta drivers). AMD expects to improve upon these results next quarter, and a variety of factors will help in this endeavor. The R600/RV6x0 products really only started to move near the end of this past quarter, and apparently the HD 2400 and HD 2600 series are doing well with OEMs and builders. Retail sales are probably lagging NVIDIA's competing products, but it should still provide some good revenue for AMD. Even though the NVIDIA 8800 series overshadows (and often undercuts) the R600 based HD 2900 XT, sales of that high end card are apparently brisk as well. So the graphics numbers for Q3 will have a much greater impact on overall revenue for AMD, and should help dig into any potential loss. Barcelona for servers will start to sell near the end of Q3, and should have a small overall impact on earnings. Barcelona for desktop will be appearing in late Q4 (Phenom processors), and will have a very minimal positive impact on revenues at that time. AMD expects to stabilize pricing on their products, and transition more of their portfolio to 65 nm production at Fab 36. AMD seems very happy with the results of their 65 nm process, and have stated that yields have exceeded expectations. 90 nm production at Fab 30 is winding down, and AMD will slowly sell off their 90 nm/200 mm wafer machinery, which is apparently worth around $800 million. With the finer geometry of 65 nm and mature yields, AMD hopes to improve ASPs while increasing their CPU sales with stronger desktop and mobile platforms based on the 690 chipset. AMD has stated that they believe that Q4 will be the break-even point for the company. Q3 will still show a loss, but if AMD's plans of cutting expenditures and growing business work we can expect a loss of between $250 million to $350 million next quarter on revenues of approximately $1.42 billion. Graphics will likely play a large part of that growth, as AMD should only make one to two percentage points of gains in overall marketshare. AMD is hitting the mobile sector quite hard, and because they have chosen multiple partners instead of an "all in one" branded Centrino type product, many laptop manufacturers appreciate the choices they have, which allows them more flexibility in their product mix and inventory control. In the second half of this year we can expect the AMD 780 and 790 chipsets to be released. The 780 could very well be the first DX10 integrated part to hit the market, and considering how small the die size of the current HD 2400 is, we can assume that much of that functionality will be included in the 780. I would suspect that the shader count will drop to 20 stream units (4 x 5sps) in one to two SIMD units. Color and texture units will likely mirror that of the HD 2400 as well. The big use of die space will be the UVD unit, which will provide full 1080p playback. If AMD can get that particular chipset out in a timely manner, it would really help to push the adoption of their CPU products in both mobile and desktop solutions. AMD did not really comment on going fab-less for production, but they did state that it takes capacity to compete. They still have the option of building the Albany fab in New York, and have until 2009 to break ground. Fab 30 to Fab 38 conversion has been slowed down, but this is offset by how well Fab 36 is apparently performing. AMD will still use Chartered for outsourcing, but could utilize IBM fab space in the future as well. There were a lot of positives in this previous quarter, but AMD still has a long road to hoe. The next year will be very important for AMD in terms of execution. They need to get server and desktop Barcelona chips out, and they need to execute on their 45 nm process. If AMD cannot do either of these things, we might in fact see them sell off their large capital assets (such as their Fabs) and go entirely "asset-lite". By Q4 we should have a pretty good idea where they will be heading. In one last thought, I do not think it unlikely that AMD will produce Athlon X2 parts on 65 nm that will reach upwards of 3.2 GHz before the launch of Phenom. AMD has a tradition of continuous improvement when it comes to their fabrication techniques, and it certainly is feasible that they can improve their speed bins to produce higher clocked 65 nm Athlon X2s running at current TDPs. It would not be surprising to see a 89 watt 3.2 GHz X2 sometime late this Fall. Point Blank Gaming LAN in Cheyenne Apparently a small group of guys in Cheyenne, WY are putting on a LAN. If you are around the Laramie/Cheyenne/Greeley area and have an interest in some gaming, these guys have the digs for you. Seems they were able to get NVIDIA and Thermaltake to help sponsor their last LAN, so there could be a few door prizes for interested gamers. Sounds like there could be some free pizza as well. The LAN goes from August 11 at 10am to August 12 at 10am. 24 hours of gaming fun and prizes. You can check out their website here, or reserve your spot at the LAN here. July 19, 2007 Intel Posts Some Nice Numbers - Josh On Tuesday Intel released their latest quarterly figures, and they lined up very well with what the Street had predicted. The only area of concern were overall margins on their products. Margins were expected to come in at 48%, but they were in fact reported as 46.9%. Most of us think, "What's a percentage or two?" but apparently many investors do not feel the same way. The perception is that many see Intel as pursuing a price war where one is not merited. As of this week, AMD does not have a desktop chip more expensive than $179, while Intel has quite a few well above that (including several of the pricier quad cores). On July 22 Intel is going to drop prices again, and this apparently has investors somewhat angry. The general consensus is that Intel is more interested in gaining back marketshare from AMD instead of maximizing their profits on their current products. When you consider the price/performance as well as power consumption, it was no wonder that AMD now has the majority of their desktop products sitting below $149. When lining up the competing products from Intel, it becomes pretty clear that the Core 2 chips are superior in performance per clock, thermals, and power consumption. Instead of charging a slightly higher premium for these chips over the competition, Intel is slashing prices so that it undercuts AMD at every speed grade. The question many keep asking is why? Why is Intel not maximizing their profits on a superior product? Why is that extra couple of percentage points of marketshare so important that Intel is willing to sacrifice margins on products that should be making a lot more money. Many feel that AMD is holding onto their marketshare very well due to good relations with OEMs, as well as a renewed focus on channel partners. If we take a longer look at the situation, I believe I see the strategy that Intel is following. Even though they are not maximizing their profits on each chip sold, the thrust is to actually gain back marketshare in the near term. Once Intel has done that, then I am betting 45 nm production will be ramped up dramatically and Intel will offer chips with even better performance and thermals, and charge quite a bit more for those. Intel's advantage in both design and process will really come to the fore this Winter, and AMD will not have a whole lot to fight back with. With Barcelona not showing up until early fall and running at fairly low clockspeeds, plus the desktop Phenom products will not show up until nearly the end of the year, Intel looks to really bring the hammer down with their 45 nm products. At that point I think we will see margins take a very healthy increase, and the marketshare gains that Intel is trying to achieve now will pay off in increased revenue and profits when the cheaper to make and faster chips hit the market. AMD announces their quarterly results this afternoon, and the general consensus is that they will have a much more positive quarter than the one before. AMD apparently has been able to clear out a lot of inventory, and their chipset and graphics sales have improved pretty dramatically since the beginning of the year (I think the 690G could be the product that is moving the most, as it really is a tremendous integrated part). With this in mind, plus the recent influx of $1.8 billion, AMD should be in a much better position now than they were 3 months ago, or even 6 months ago. They still have a long road ahead of them, and it will take a lot of effort and some more cuts (like executive bonuses) to keep afloat until their 45 nm process comes online next summer. I think Barcelona will help keep the ship upright, and the eventual Phenom release will help as well. But until AMD can transition to 45 nm with an advanced Barcelona based core, I think this year could be rather rough for them. July 18, 2007 Intel and the 1333 FSB - Josh Several days ago Intel released the QX6850, the E6850, and the E6750 which all feature support for the new 1333 MHz FSB, which is a prominent feature of the new P35 chipset (and upcoming G35 and X38). These of course are the fastest processors you can buy on the market, far outpacing the AMD X2 6000+ (the QX6850 is a quad core chip running at a blistering 3 GHz, the E6850 runs also at 3 GHz but is dual core, and the E6750 is running at a not-so-sedate 2.66 GHz). Intel seeded very few of these (comparatively) to review sites, as the performance falls in line with previous chips at those (or close to) speeds. Power consumption is also very similar, even with the faster bus. I was originally part of the "The FSB is dead" group, warning that performance on Intel chips would be drastically reduced or inhibited by the use of the aging FSB that was originally based on the Pentium 4 chip. I must admit now, the bus isn't holding Intel back. Even with a quad core chip, very little performance difference exists between chips running at 800 MHz FSB and 1333 MHz FSB. While I still feel it is worthwhile to get off the bus, Intel certainly is not at a point where it needs to. AMD just plods along, improving their 65 nm process, and exploiting the only potential advantage it has. Price/performance and power consumption are the factors keeping AMD from being totally overshadowed by Intel. While things seem dim for AMD, we can only look back and see that things have been much worse. Consider how many years the K6 struggled against the Pentium II, or how the Athlon XP could just not keep up with the Pentium 4 Northwood cores. When we look back at those times, we see that factors such as motherboard support and chipset quality were also negatives held against AMD. Intel always seemed to be the solid choice, with great reliability with their chips, motherboards, and chipsets. Now we see things as quite a bit different. NVIDIA has some great chipsets (though getting long in the tooth right now, but still quite competitive), AMD is releasing their own chipsets based on the ATI parts, and VIA and SiS still release products supporting them as well. It is quite a difference when compared to previous generations. The near future is still not terribly rosy for AMD, as they are quite some time away from releasing their Barcelona based processors in the Opteron market. They are farther still from releasing desktop parts based on that core as well. Rumor has it that the chips will see a max speed of 2 GHz at release, and very slowly will increase to around 2.3 GHz by the end of the year. Even with the potential architectural advantages over the competing Core 2 products from Intel, it does not seem that AMD will be able to crank up the clockspeed enough to overcome the performance of the current 65 nm Core 2 parts, much less the upcoming 45 nm Penryn based products. The next year could still be quite rough while AMD ramps production of the Barcelona variants, and I would not expect ASPs on their parts to be much higher than they are now. It will not be until AMD releases products on their 45 nm process, plus putting the time into that particular variant of the K10 architecture, that we may see the performance crown swing back to AMD. One year from now is quite a long time in the semiconductor industry, and a lot can happen between then and now. Still, we have to congratulate Intel for the fine job done with the Core 2 products. With the price cuts coming up this next week, we will see the Q6600 literally fly off the shelves at the $266 price point. AMD has already lowered their prices, and it is doubtful they will lower them again. Things are going to be tight for the little green machine, but at least their HD 2600 and 2400 parts seem to be popular with OEMs. That should help the bottom line a bit. This fall will be a busy time for AMD, as they will be releasing their 65 nm high end graphics chip, as well as the Barcelona/Opteron launch. Until that time, Intel will be sitting in the catbird's seat. July 11, 2007 News Around the Industry - Josh Yesterday's announcement that Dave Orton was leaving AMD came as a surprise to many. Dave was one of the prime movers of the AMD/ATI merger, and as such he was still basically in charge of the company formerly known as ATI and reported directly to Hector Ruiz. Conspiracy theories abound about his departure, especially in light of the seemingly unimpressive HD 2x00 series of launches. Dave himself has basically said that he was extremely tired of the commutes that he has had to endure for the past seven years (he lives in California, ATI is headquartered in Toronto/Markham, and AMD's main offices are in Austin). As you may or may not remember, Dave was in charge of ArtX before ATI bought up that little company. ArtX was primarily the team behind the R300/R350 chips which dominated the high end graphics industry for a few years. While ATI has not had another breakout product like the Radeon 9700 or 9800, they certainly haven't done poorly. The X800 series were competitive with what NVIDIA had at the time, and the X1900 series was the most feature packed and performance leading product in the market until the 8800 series from NVIDIA came to town. The HD 2900 XT was late, and it has some interesting "quirks", but it is for the most part competitive with the NVIDIA products in its price range. The midrange HD 2400 and HD 2600 products are apparently selling quite well in OEM circles, as they perform well for their price range and their power requirements are very low (cooling requirements are also quite low as well). So overall Dave is leaving the graphics and media division of AMD in an interesting place. Their new architecture is not unimpressive, but the halo effect of having a dominating upper midrange and high end set of products is holding them back. What is missing so far are a set of DX10 supporting products spanning from the $149 to $349 market, which is pretty significant. NVIDIA has the 8600 GTS and 8800 GTS 320/640 products slotting nicely into that market, while AMD has the aging X1950 Pro and X1950 XT sitting there to compete. We can certainly hope that this coming Fall the high end R600 refresh will be a bit more competitive at the high end, and that will likely drop the current R600 based cards down in price to address that large gaping hole in ATI's lineup. In other news, Damage from Tech Report has "finally" released his HD 2400 and HD 2600 review. A full one this time. With commentary. And benchmarks. July 3, 2007 BFG Unveils Thermointelligence - Josh BFG announced yesterday the release of their Thermointelligence series of GeForce 8600 cards. No one has yet been able to dissect one of these cards, but according to BFG the heatsink and fan design are significantly better than the stock options available to most 8600 cards. The first two cards to get the Thermointelligence treatment are the 8600 GTS OC and the new 8600 GTS OC2. The OC2 sports some nice clocks off the bat. The stock 8600 GTS runs at 675 MHz core, 1450 MHz shader clock, and 2000 MHz memory. The OC2 ups those to 720 core, 1566 shader, and 2200 memory. This should give a pretty healthy boost to overall performance over the stock product. The extra shader speed should come in really handy as well! No word on how quiet this product is, but it should not be any louder than the stock cooler. The numbers that BFG claims are actually pretty impressive. Upwards of 11 degrees C cooler in SLI than the stock coolers. Single cards can show upwards of 5 degrees C cooler as well. Since NVIDIA will likely not refresh their midrange cards until Spring of 2008, it is nice to see some more separation of SKUs from the different manufacturers. Choices that actually rely on unique features helps the marketplace a lot, as compared to just choosing identical cards from different brands. BFG provides an adequate bundle for their card, but nothing terribly exciting. I am curious if they will include their unique case stickers. The OC2 card is already available on Newegg, but the price is pretty staggering at this time. While most 8600 GTS have gone down in price significantly, the 8600 GTS OC2 is at the full $239.99 Other Interesting News It seems that the average prices of the HD 2900 XT have finally gone below the $399 mark originally set by AMD. This is a nice break for consumers, as the $419+ prices were getting somewhat annoying. Something else of interest is the release of the 1 GB cards from Diamond and Sapphire. Currently the Diamond card is only available to system builders, but I noticed that Newegg does carry the Sapphire 1 GB card. The price is a lot more reasonable than I expected to be honest. $489 is still far less than the lowest priced 8800 GTX, and performance of the HD is really not all that bad. In some tests it does seem constrained by the 512 MB of memory the standard version has, so this extra memory space may help out quite a bit. The latest 7.6 drivers help overall performance and smoothness quite a bit as well, not to mention fixing quite a few bugs. A lot of people were expecting a whole lot more out of the HD 2400 and 2600 launches, and many of the initial reviews were quite negative. I can see where most are coming from, since a lot of people expected the 2600 XT to outperform the 8600 GTS. That was obviously not the case. When we take a step back and look at the marketplace though, we see a lot more parity than first impressions gave. The HD 2600 XT will cost around the price of the 8600 GT, which it does outperform by a good margin. Again, AMD has placed these products in price ranges that are competitive with NVIDIA's cards in terms of overall price/performance. So it is not all doom and gloom at AMD, but I think many enthusiasts (vocal ones) are disappointed that AMD/ATI has not been able to hit one out of the park for several years now. Who knows what to expect this August/September though. The 65 nm version of the R600 should be seeing the light of day, and many are hoping that this will be the high end savior for AMD Graphics. I hate to pin the hopes of an entire company on one product, but it certainly would be nice to see the performance needle swing back again to the AMD side (even for just a little while). Competition is a wonderful thing, and I certainly do not think it would hurt NVIDIA to be second best for a while. When we consider that the X1900 XTX and X1950 XTX were the fastest single chip cards on the market for quite some time, it certainly did not hurt NVIDIA's bottom line. I think that the biggest single improvement with the 65 nm product will be the performance cost of shader based MSAA resolve. I do not think AMD will radically change the Render Back Ends to include fixed function MSAA resolve, but I believe they will certainly improve how it works with the shaders. Time will tell though. June 28, 2007 HD 2600 and 2400 Day - Josh AMD has officially announced the release of the HD 2600 and 2400 series of products. These are some of the first 65 nm mass produced graphics products to hit shelves, and they should be available around the first or second week of July. I have been looking forward to this release, as I expected the performance of the cards to trump that of NVIDIA's latest 8500 and 8600 products. I also expected them to come in a bit cheaper than the NVIDIA products. I was obviously half right. The top of the line HD 2600 XT will show up around $149, which is less than the unofficial list price of the 8600 GTS ($179). One would expect performance to be a little bit better than the 8600 series from an architectural point of view, but that is not the case. The HD 2600 XT simply cannot overcome the 8600 GTS in a large majority of applications. Sure, there are a few where it does overcome the performance deficit, but they are few and far between. One area that is interesting is in DX10 performance, but the overall rates for either card are far less than playable in the currently available DX10 titles and benchmarks. Things get even worse when AA is enabled, and performance really takes a hit. This is probably two-fold. The first is that ATI again designed a product with only 4 Render Back Ends (RBEs). This is against NVIDIA's more traditional ROPS which are double that. So there is a major fillrate difference between the two. The 8600 GTS is clocked at 675 MHz, while the HD 2600 XT is at 800 MHz. This also means there are double the amount of multi-sampling units on the 8600. The second issue is that of MSAA resolve. In NVIDIA's hardware it is designed to both support DX10 specifications in that it can do MSAA resolve on the shaders, but it also has fixed function resolve for applications which do not require shader MSAA resolve. While shader based resolve is not a bad thing, it certainly is not as efficient as fixed function circuitry handling those jobs. In my HD 2900 XT review, I will be going over some of the reasons for this, as well as if this is a design decision or a hardware error. The good part of this release is the price. The cards are performing where one would expect considering the prices. We also have the addition of the UVD (Universal Video Decoder) which is a step up in functionality from NVIDIA's PureVideo 2. Another good thing is the power draw of these cards. The high end HD 2600 XT pulls around 45 watts at max, which is quite a bit less than the near 75 watts the 8600 GTS draws. Apparently these cards are very popular with OEMs as their price, power draw, and feature-set fits nicely within a multitude of consumer systems. 3D performance is not class leading, but most major OEMs do not care about that (look at how many Intel Extreme Graphics parts are out there). It is good that AMD was finally able to release these parts, and they will likely make some good money off of them with OEMs. Retail sales may be a different ballgame. NVIDIA has some awfully deep roots in the retail market, and the performance and feature-set of their 8600 parts overshadow that of the HD 2600 series. Now that prices of the 8600 GTS products are approaching $150 (with rebates), I am sure we are going to see a lot more movement of these parts. From an enthusiast point of view, this was a miserable failure. From a business point of view these products will probably sell in bucketloads. ATI has focused on math performance rather than more traditional pixel and texture fillrate for the past few generations. I personally feel that this was the wrong direction for them, as the majority of games out there still require a lot of traditional fillrate. Not my call though, and these design decisions were made years ago when the promise of shader heavy DX9 titles as well as a slew of mature DX10 titles were expected. Little did those planners know that Microsoft would continually delay the Vista operating system, as well as DX10. When we look at this release, we invariably look to the future when AMD will release the 65 nm refresh of the R600. Will that be the high end savior of AMD like the X1900 was a year and a half ago? Or will it still show the dramatic performance drop of using AA due to shader based MSAA resolve? Time will tell, and we will find the answer early this Fall. Until then, we have the new 65 nm midrange products from AMD to play with, and hopefully extract a bit more performance out of. June 20, 2007 NVIDIA's Tesla Computing - Josh NVIDIA is taking the bull by the horns and releasing three standalone products that support their GPGPU initiative. Based on the G80 chip, these three products span from the single card to a full 1U rack packed with GPUs. While the basic board is reminiscent of the 8800 GTX, these cards feature 1.5 GB of memory and over 500 gigaFLOPS of processing capabilities. This could seriously open up a new avenue of income and customers for NVIDIA, plus have a very positive effect on High Performance Computing everywhere.
You can read the entire article here. June 12, 2007 Auzentech and Other News - Josh Two pieces of Auzentech news this morning. First off they have released their DIN upgrade the X-Tension. This little number enhances the I/O capabilities of the current X-Meridian and the upcoming X-Fi Prelude. Prices should be forthcoming, as well as availability. As you well may know, I really enjoyed the X-Meridian product, and think that it is one of the most capable soundcards one can buy for the money. The addition of the X-Tension DIN will only enhance that reputation. The second piece is that the X-Fi Prelude should be available for sale by the end of this month. This is perhaps one of the most anticipated sound cards in recent history, and the first 3rd party standalone soundcard based on the X-Fi chip from Creative. AMD is shipping HD 2400 and 2600 products to their partners, and we should see boards available by the end of this month. Quite a bit of controversy surrounds these products. Some are saying that both will be faster than the NVIDIA options, some are saying they will be significantly slower, and then there is a lot of speculation that falls between these two extremes. My personal take is that the HD 2600 XT will be faster overall than the GeForce 8600 GTS, but who knows how the HD 2400 series will perform? Perhaps all I care about is that there will be new products and more competition at the very important $159 to $249 space. The use of GDDR-4 with the HD 2600 XT could really help push it over the performance of the 8600 GTS, which still uses fast 1 GHz GDDR-3. Reviews should be popping up closer to release (think two weeks). NVIDIA has slashed the prices on the 8800 Ultra, and I have been told that before this cut that "demand has been soft" from a partner. $829 was a lot to ask for a card that was only slightly faster than the $549 8800 GTX. Now we are seeing prices reach well below the $700 level, which is a bit more reasonable. Other than that, we haven't seen much movement in prices as of late. The 8800 GTXs have stood around $549 for the stock version, slightly higher for overclocked versions. The 8800 GTS 640 is around $329, the GTS 320 is around $279, and we have seen the 8600 GTS parts fall well below $189. All in all, a decent time to spend money on a graphics card, because I really don't foresee anymore massive changes in price until at least a early September timeframe. At that point we would expect to see the 65 nm refresh of AMD's R600, which could very well challenge NVIDIA for the top spot. One final bit of news is that it seems as if we will in fact be seeing a 1 GB HD 2900 XT in the next few weeks. This product was supposed to be available at launch, but due to the overall performance and price such a product would show, ATI/AMD thought it better to skip for the time being. Well, it seems some of the partners want to in fact release the unit, and we have seen the first pre-orders pop up on the web. I think overall this is a good thing, as I have experienced some titles which did seem like they needed more than 512 MB of memory, especially at higher resolutions and AA levels. Two of NVIDIA's products have more memory than 512 MB, so it really is about time to offer another product with more juice. Since it will be using the same R600 chip and similar core clockspeeds, we should not expect any miracles in the performance department.
If you have found this article interesting or a great help, please donate to this site.
Copyright 1999-2007 PenStar Systems, LLC. |
|||
