   |
|||
|
|||
|
:
News
: Archives
|
|
|
|
|
|
|||
|
|
|||
|
|
|||
|
July 29, 2005 A Slow Month - Josh And I'm not just talking about tech releases! I essentially took this month off to get a few things done in the real world. This included getting a refinance from Hell done, as well as taking some time off from the industry. I must admit that I am feeling refreshed about things, and I have a real urge to write again. I have a slew of reviews coming up, and perhaps one of the more interesting things is that it looks like I will be doing some freelance writing for the folks at Extreme Tech. I am quite excited about this, and hopefully I am up for the challenge. The guys writing at Extreme Tech now are a very talented group of individuals, and I am quite honored to be able to write with them. I will continue running my site, and this will only be a freelance position (one to two articles a month, depending on their budget). Still, I think it is going to be an exciting opportunity. I have been thinking more about SLI-AA as of late. First off, I think it is becoming very clear that it is a very performance intensive application. We are seeing some pretty hefty performance hits with many current applications, and running at 1600x1200 SLI-8X just won't cut it for many games. Some older games, yes. Doom 3 and the current crop of visually stunning games, no. It is a bit disappointing that the performance just is not up there as much as one would expect, but I really don't see that we have much to complain about. When SLI users originally bought their setups they were not expecting this feature to be added, so it is essentially a freebie. Still, there are applications which will benefit from SLI-AA, especially older titles which may not play nicely with other SLI and AA modes. Yesterday I chatted with Nick Stam about my idea that some titles which never played nice with AA (Halo, Splinter Cell, etc.) might in fact be able to utilize SLI-AA to get at least the supersampling advantages to work. Again, the NV guys are working on this one to see if it will go anywhere, but the possibility is there that it might just work. I guess we shall see. So, while SLI-AA is not the panacea that many of us were hoping for, it is another nice visual quality option that we can enable. I must admit I am very, very curious how ATI's Crossfire Super AA will work. This could be an area that ATI has a performance advantage in due to the nature of the compositing chip, but we won't know that until we see official reviews with full feature functionality. Speaking of ATI, it appears that August will be the month where we will actually see Crossfire boards hit the market, and the introduction of the R520. I would expect the R520 derivatives to be released in September, and allow the R520 to take the thunder early. I doubt NVIDIA will stand idly by, and we will see the release of the 7800 GT in early August, as well as the possibility of seeing a 7800 Ultra type product to counteract the R520. Due to NVIDIA's ability to get working silicon into its partners' hands, NVIDIA is definitely in the drivers seat for this generation of products. There are also rumors of midrange parts that are ready to counteract anything ATI has coming up. So, needless to say, we will see a lot of action from both ATI and NVIDIA this fall. I love the competition, it certainly keeps us all on our toes! July 22, 2005 As I was surfing around the web, I noticed a couple of interesting editorials. These two are from Sander and Ed. If the only reason for these editorials is to generate hits, they certainly have done that. Their logic and reasoning however... Anyway, Ed goes off about what seems to be a plethora of new sockets in the future for AMD. I can understand his thinking behind this, but I don't think he truly grasps the reasons why AMD has done what they have If there is any processor company out there that understands the advantages of developing a long lasting socket technology, it is AMD. Socket A was around for ages before finally being phased out. This made end users, motherboard manufacturers, and OEMs quite happy. I honestly don't think that socket 754, 939, and 940 were poorly thought out either. Yes, AMD should have probably introduced 939 first, and kept 754 mobile, but where exactly would that have left the Semprons? I think the progression that AMD has had with initially offering 754's and 940's was actually quite logical, especially since the 754 design was a lot simpler to implement on a motherboard, and because of the single memory channel it would fit into laptops a whole lot better than the dual channel 939's. By the time that the 939 socket will be replaced, it will have been around for 3 years. Socket 754 will have been around for almost 4 years at that point. I don't see how these are really considered short lived. The next issue that Ed brings out is that the next gen of sockets will be the M1 (638 pins) and M2 (940 pins, but not compatible with current 940 pin processors). I am sure that AMD would have loved to keep the same sockets for these products as are currently offered, but they simply can't. This is due to the upcoming support of DDR-2. The last thing AMD wants to do is offer an identical socket for DDR-2 support as DDR-1. How many burned out processors would come back to AMD because some user or OEM accidentally installed a DDR-2 Athlon 64 into a DDR-1 motherboard? I think the situation would sow so much confusion, that it is definitely not worth it for AMD to try such a trick. So, while Intel has seemingly sidestepped this issue by continuing to use a traditional Front Side Bus, they also have a plethora of sockets to choose from (478, 479, 601, 602, and 778) for their different products. In the end, I really don't think AMD made any real mistakes with its socket support. 754 is for mobile and budget desktop, 939 is for mainstream and performance desktop, and 940 is for Opteron. I do admit that AMD and its partners might in fact start blurring the lines a bit, especially in regards to offering Opterons in 939 packages. That is not so good, but I can understand AMD trying to implement lower cost servers. Still, AMD does not make these socket changes on a whim. Sasser on the other hand... his article is basically saying that AMD's marketing is at fault because they don't send out enough samples or have enough of a television or print presence. For a company that makes $1.2 billion a quarter, and barely pulls a profit in any quarter, marketing is not the highest priority. Should it? Probably, but any company with a limited budget has to divvy up the money as best they can. Intel spends billions in marketing every year. They probably spend a good $5 billion a year in getting the Intel presence known. That is more money than AMD makes in a year, total. So, AMD does what it can to have a good presence where people will see them. They support the Ferrari team, and right now they are big with Lance Armstrong and the Tour de France. These two things do give them a higher profile, but that has essentially drained the marketing budget. Also, as for sending out several hundred processor samples... that is ludicrous. Intel doesn't even send out "several hundred" samples for their launches. Samples are essentially lost revenue for these companies, and even though there is a tax write off for them, it is not smart for any company to give away hundreds of expensive parts simply for mollify a couple of hundred reviewers. Let's say AMD did send out 200 processor samples to different publications. Sure, they would get a lot of reviews, but how effective is that money spent? How many people are going to go out there and read all 200 reviews? So, while AMD has provided all these samples, and probably gained a few readers, the efficacy of each sample sent out is essentially diluted. So while overall readership of this new product might increase by a couple of percent, AMD has spent a LOT of money to achieve a very small gain. 200 FX-57's would essentially be around $200,000 of lost revenue. How many engineers, clerks, accountants, marketing personnel could have worked at AMD with the money that was lost by sending out so many free review samples? Consider that AMD has about three big releases a year. That's around $600,000 worth of lost revenue. When AMD is pulling a couple of million in profits each GOOD quarter, every little bit helps. If they can increase their productivity by hiring 6 good engineers at $100K a year a piece, I think they would rather do that then flood the press with free review samples. Sure, we all like to get free products, but I think AMD has thoroughly thought this one out, and the samples they do send out find a large reading population as is. I know when a new processor comes out I read on average about three articles pertaining to that. I think AMD does around 50 to 75 samples of any product, and I know that I don't go out and read every one of the reviews on them. Both Intel and AMD are doing the best they can in terms of technology and marketing (albeit Intel pretty much has the marketing side down pat), so this talk of confusing sockets, not enough samples, and not having their cake and eating it too is just kinda silly. July 20, 2005 Forceware 77.76 Preview - Josh Today NVIDIA released their beta 77.76 drivers, which are currently awaiting WHQL certification. While still fixing small bugs and problems, NVIDIA decided to add a lot of functionality for the SLI crowd. Now NVIDIA officially supports the 6600 and 6600 LE PCI-E video cards in SLI use, as well as adding two new SLI-only antialiasing modes. Here is a quote from the preview:
You can read the entire article here. July 18, 2005 News Around the Industry - Josh It seems that there is a small flurry of new drivers hitting the web. First up is ATI's Catalyst 5.7 drivers. These have a couple of fixes to some applications, but it has two pretty nice features. The first is performance. In quite a few apps 128 MB cards will show a significant gain in performance due to optimizations with the driver's memory functionality. For those with the smaller framebuffer cards, this will be nice. The second is the addition of 1080p HDTV rendering. For those HTPC enthusiasts with ATI cards, this will be a welcome addition. NVIDIA released its AMD nForce 4 only 6.66 drivers (yes, that is a nice number). These features some general compatibility fixes, but no word if it takes care of the networking issues that the Tech Report has uncovered. Quite a lot of talk has been going around about ATI's R520, most of it coming from The Inq's Faud. There are a few things that I have gleaned as of late that I can share. It doesn't appear that ATI will be unveiling the R520 on the 26th of this month as we were expecting. Apparently NVIDIA's 7800 GTX release and instant availability made a significant impression on the folks up north, and I think we will be seeing the actual release sometime in September with limited availability. It seems the final respin that took place earlier this summer is doing the trick for ATI in terms of yields and speed bins. We can expect clock speeds of the R520 to hit above 600 MHz for the top end model, and this speed I believe is what caused the majority of problems for ATI. Many believe that the R520 will only have 16 fragment pipelines, but anymore what exactly is a fragment pipeline? My personal belief is that each pipeline will contain quite a few functional units to it, so they will be very efficient pipes. Clock speed will also be an advantage for ATI over NVIDIA, but I think overall the two companies will have products that will compete well with each other. There is no word from TSMC though about ATI ordering production levels of wafer starts on their 90 nm process, so I seriously wonder what the availability will be like upon release. My guess would be limited availability on the day of the release, but supplies will dry up quite quickly. It also seems that NVIDIA is in fact holding out on an Ultra edition of the 7800. It appears offhand that the chip will be comprised of a higher clocked G70 (110 nm process). Many are estimating 500 MHz core and 650 MHz memory (1300 effective). The 7800 GTX that I have in my possession runs quite nicely at that speed, and it does show a hefty speed boost from it. Now, with all the controversy of the three internal clocks that have been discovered in this card, I am guessing that NVIDIA will change around the internal clocking scheme as well to make this much more effective than a simple "clocked up" 7800 GTX. NVIDIA is of course waiting for ATI to release the R520 before unloading this beast on us, but we can expect the Ultra edition to probably feature 512 MB of memory. By the time September rolls around, we can expect the 7800 GTX to fit nicely in the $499 range, while the Ultra will be at the $599 price point. I do not expect NVIDIA to move their high end products to 90 nm until at least late November, but there are some kicking around the idea of a January/February timeframe for another high end release. What will change though is the midrange and possibly budget products. Apparently TSMC's 90 nm process is very robust at this time, and NVIDIA sees a definite crossover point at this time for the mainstream and budget products. G71 looks to be the midrange refresh, and looking at the falling prices of the previous midrange champ (the 6600 GT), I would expect this part to be a full 16 pipeline G7x design with probably 8 ROPS running at around 500 MHz, all the while producing very little heat and pulling very little power. If NVIDIA decides to get really aggressive with this design, then we would expect to see clockspeeds around 600 MHz. It will probably feature 256 MB of memory, but the big question will be if they utilize a 256 bit memory bus, or keep things simple at 128 bits. Either way it will be a solid performer for its pricerange. I think the big question is what NVIDIA will think ATI will do, as ATI already offers the very solid X800 with a 256 bit bus that costs less than $250 these days. The G71 will be about the same die size as the current NV43 (6600 series), but will significantly outperform it. The G72 looks to be the budget offering, and I can seriously see this product going one of two ways. The first is that its functionality is not overly increased from the NV44, in that it features 2 ROPS and 4 fragment pipelines (plus all the other video features) but it is clocked amazingly high as compared to the NV44. We could expect clocks to easily hit 500 MHz (stock NV44 is at 350 MHz) all the while keeping the cooling on this chip extremely minimal. If NVIDIA has not increased the functionality or complexity of this chip in any extreme manner, then we can expect an incredibly small die size, which means that on 90 nm with 300 mm wafers, we could expect die sizes around 64 mm square (which is very small) and on a 300 mm wafer there could easily be 1000 gross die. Now, not all of those die would be good, but a conservative estimate would see around 800 of them as good. I am not sure how much each finished wafer costs, but it should be between $1,500 to $3,000 per finished wafer (not counting cutting the dies out or packaging them). So, between $2 to $4 per raw die is damn good for a budget product, add in around $5 to $10 for further processing and packaging per die, and we can start to see some very good performing yet cheap parts later this fall. Anyway, from all appearances both ATI and NVIDIA will be releasing mainstream and budget parts at 90 nm, but we of course have heard very little about the functionality. Both companies could embrace the high end by offering a 16 pipe midrange part clocked aggressively, along with a 8 pipe budget part also clocked high. Or we could see them keep their budget products as 4 pipe solutions and keep the clocks on all their products fairly conservative. I guess in the end it will come down to yields and speed bins, and how much die space each company is willing to sacrifice. Interesting times in this industry, as usual. July 14, 2005 AMD Posts 2Q Results - Josh In a bit of a shock to Wallstreet, AMD posted a small profit of $0.03 per share, on revenues of $1.26 billion. While most analysts expected AMD to post a small loss, the strong processor sales and a declining weakness in memory sales helped AMD over the hump in what is typically a very weak quarter. AMD is planning on spinning off its Spansion memory group, and if that happens then the overall positive outlook for AMD will improve. While in the past the memory group has kept the company afloat during lean times with AMD's processors, they see the memory products becoming a commodity market, where margins are non-existent. The only problem with this is if memory prices and margins improve, and AMD has problems selling processors again. AMD is reporting processor sales up a pretty astonishing 38% over 2Q 2004, so year on year growth has been pretty phenomenal. AMD still contends that Intel continues to put significant roadblocks towards greater adoption of AMD parts. Whatever the case, AMD is doing well with its processor sales. ASPs are probably well above $100 now, especially when one considers that even the lowest X2 processor is more than $500, and the 800 series of Opterons are above $1000. The total conversion to 90 nm products has helped AMD immensely with their average cost per die, and I think that is really reflected in this quarters performance. AMD expects 3Q to be stronger than usual, so we can really expect the processor group to make a very healthy profit. And with the way that Spansion has handled this previous quarter, I think that they will again be healthier than most analysts expect. While they will not pull a profit, I would expect them to cut losses by at least $13 million (if this trend continues). With increased processor sales and profits, AMD could be in a good position to report a $20 million + profit for Q3 with revenues approaching $1.31 billion. While this of course does not compare with Intel's expected $2.2 billion in profit on $10 billion in sales, it is a good step forward for AMD. A Controversial Thread - Josh I just had to post this. Perhaps I am a bit extreme in my stance, but I just have to shake my head on what the media is reporting on, and what politicians are complaining about. Opinions are very welcome, but try to keep insults to a minimum! July 13, 2005 An Amazing Lack of Updates - Josh Refi's really stink. We were supposed to close at the end of June, then the first week of July, then we had an official closing yesterday. But of course it hasn't happened yet. Apparently some paperwork was not signed off when it was supposed to, and now we find out the title company doesn't have another opening to close until sometime next week. This just keeps getting better and better. Add on top of that the University of Wyoming, for some messed up reason, decided to pay my wife 1/2 of what she is supposed to make this summer. I'm hoping that all this garbage decides to end soon, because honestly, enough is enough. Needless to say the urge to write and test while trying to straighten out this entire situation has been pushed far into the background. That is unfortunate, because June was one of the best months ever for the site. I was hoping to keep that ball rolling, but oh well. I might be doing some freelance work for Extreme Tech in the next coming months, but some details of that need to be worked out. I was seriously considering joining that group, but it involved moving out to San Fran. With a family and a wife trying to finish a degree at the University, it just turned out to be too much of a change. Still, this does give me the chance to really grow with this site, and I certainly hope to continue that. Now some interesting things around the web. The Tech Report takes a look at high resolution gaming (above 1600x1200), and they came across some really interesting results. It appears that most current generation cards really only handle the 2 megapixel range well, and their compression and Z-cull operations are really tuned to that. Once 2 megapixels are exceeded, then performance on most of these cards takes a nosedive. Even the 6800 Ultra SLI setup takes a massive performance hit. The only exception? The new 7800 GTX has been optimized so it can handle up to 3 megapixels. This means that it doesn't take the dive once 1600x1200 is exceeded, and keeps strong until the maximum resolution of 2048x1536. 7800 GTX SLI is of course unbeatable, and the results of that combination are pretty amazing at that resolution. Read the entire article here. HardOCP takes a good look at the BFG GeForce 7800 GTX, which many consider the cream of the crop of reference cards. BFG packs a little bundle with the box that is definitely not common, but the card comes out of the box overclocked by 30 MHz (460 MHz vs. the standard 430 MHz). Oddly enough their sample didn't overclock extremely well, but it certainly is still a great card to have. Chris over at AMDZone has a review of the Albatron K8SLI, one of the smallest and most inexpensive AMD/SLI boards in existence. I am hoping to get one of these numbers at the end of the month for review! Anandtech gets their hands on a reference ULi M1695 motherboard, which natively supports PCI-Express and a true AGP 8X slot. The design is somewhat unconventional, but extremely flexible. ULi has pulled a strong contender out of their pocket, and with their increasing marketshare and revenue due to providing the southbridge to quite a few different products (mainly the ATI chipsets), ULi is definitely a company to watch out for. Hondaman, a friend of mine for the past few years, is selling his BFG GeForce 6800 Ultra OC AGP model. He is only asking $300 for it, and it would be a great bargain for those looking to upgrade on the AGP side. That's all I have time for now, but stay tuned as more stuff will be coming this way! July 7, 2005 Leadtek WinFast 7800 GTX - Josh Leadtek has announced their entry into the 7800 GTX world with the WinFast 7800 GTX. I currently am in the middle of testing one of these cards, and I can attest to the quality and functionality of the product. While based on the reference design, Leadtek enables VIVO functionality (something not every 7800 GTX has), and it packs a pretty hefty punch when it comes to the bundled software. The games include Splinter Cell: Chaos Theory (brand new game) and Prince of Persia: The Warrior Within (yet another new game), as well as video and DVD editing and burning software. It also features HDTV out, which could make this product quite interesting for the HTPC crowd. It of course comes in at $599, which is the base price for nearly all 7800 GTX's. Leadtek did manage to up their warranty to 18 months, which is a step above other manufacturers (but still not as nice as BFG's warranty- at least in the United States). You can find this card right here. Terrorist Attacks We would like to send our sincere condolences to the victims of the terrorist attacks in London. While these terrorists might have influenced the elections in Spain with their bombings, I think that expecting the British to do the same is a bit ludicrous. The UK will stop at nothing to catch these people, and it has probably strengthened their resolve to fight terrorism abroad. Keep that stiff upper lip! July 5, 2005 AMD and Intel Thoughts - Josh I hope everyone in the US had a safe and happy July 4th, and I certainly hope those outside of the US had a nice time as well! AMD is continuing on with its case against Intel, though some of my info was not quite accurate in the previous post. First off, it isn't the Japanese courts that are behind AMD, it is their version of the FTC. However, AMD did officially bring suit against Intel for $50 million in damages in Japan, but that case is of course pending as it was just submitted last week. There appears to be a pretty significant split between companies that are standing with AMD, and those that are doing their best to stay out of it. Dell and Toshiba are probably the two biggest that are trying to stay away from the discovery process, but those two of course are the biggest Intel only companies out there. AMD will get what it can from them, but one wonders how much "evidence" was lost or deleted. My gut feeling in this, and mind, I haven't read through the entire brief, is that in the end Intel might have to curb some of its marketing in relation to money given to Intel only firms, but that may be the extent of it. We may hear of a few deals that were illegal, but it definitely won't sink the Intel boat. AMD might get $50 million from the Japanese suit, but I highly doubt AMD will file the same kind of thing here in the US. I could be wrong of course. So, in a year and a half to two years from now, we will finally see what the courts decide here in the US. But I don't expect any significant change in Intel's way of doing things. What I do expect is that in the short term (until the case is closed) we will see Intel be very careful with its marketing dollars, and how it interacts with the major OEM's and dealers. The timing of this was not random at all. AMD could have a banner year in terms of gaining marketshare, and by applying this very clever legal pressure to Intel at this time, they could stand a chance to gain a lot more sales. While this discovery period is going on, Intel is going to watch its P's and Q's very carefully. While I have not always agreed with the way Intel does things, it is not necessarily illegal, but with the specter of this lawsuit hanging above Intel, they are not going to be going after their customers nearly as hard as they once did. This will help AMD open the door to some new OEM's, and probably give current OEM's a better chance of buying and selling AMD based products. AMD has a technology advantage over Intel in the desktop and server market. There really is no denying that. The Athlon 64 and Opteron products are truly top notch, and the new dual core products take that performance up to another level. On the notebook side, the Turion is pretty close in performance with the Pentium M/Centrino products. Centrino is a bit better in terms of power, but overall they are very comparable in speed. One area that is lacking on the mobile side is that of integrated video. The Taiwanese chipset folks have integrated Athlon 64 chipsets, but these are not very popular. ATI is really starting to take a good chunk of this market with their excellent Radeon Xpress 200 series. NVIDIA is also readying their integrated C51 chipset, which uses GeForce 6200 level graphics. AMD is going to have to be very aggressive with the Turion products to get them into the notebook market, but with these two chipsets it stands a much better chance. I by no means do not think that AMD is totally out of whack here, as I distinctly remember my first Comdex (1999) when the Athlon was introduced. I went to quite a few motherboard companies to check out their products, and whenever I asked about an Athlon board I was shooed into a back room and shown it under wraps. These folks were distinctly scared of showing these boards in public because Intel apparently had people wandering around and would report if the motherboard folks were showing Athlon based products. These motherboard people could show as many K6 supporting products as they wanted, but Athlon was off limits at that time. I think that instance definitely flaunts the idea of legality. I also remember the infamous Computex where Intel forced their motherboard partners to cut away their VIA balloons, as the VIA Apollo Pro 133A was outstripping the i820 chipset of sales at that time. This of course happened in Taiwan, so it may not be applicable in US courts. AMD argues that with their technology advantages at this time should mean far more sales, and they should be gaining significant amounts of marketshare. So far this has not been the case. While they have been slowly taking some marketshare away from Intel in the desktop, AMD expected their Opteron to take off like a rocket. So far that has not been the case. While AMD is selling more server chips then ever before, they are not as widely available as one would think. This is the area the Xeon controls. While there have been some big defections (HP offers quite a few Opteron boxes), and some huge changes (Sun adopted Opteron for its servers), AMD has not gathered any real momentum in this very important market. AMD is certainly hoping that this lawsuit against Intel will certainly pry a few new doors open. I am certainly not suggesting that Dell will have an Opteron server anytime soon, as they will stay very close to Intel even through all of the legal discovery process. What it very well may do though is allow companies like Gateway to go back to offering AMD products, not to mention allowing other biggies like IBM and HP to offer a larger range of AMD products on both the desktop and server sides without fear of Intel imposing higher prices for the Pentium/Xeon processors or delaying marketing payments because of a larger AMD product mix. The AMD Athlon 64 performs as well, if not better in most circumstances, then the Intel Pentium 4. Not only that, but its power draw and heat production are well under that of the closest competing Pentium 4. This is actually quite huge for OEM's. It costs a lot more to design and ventilate a Pentium 4 based machine than it does for an AMD Athlon 64. Not only that, but to keep it silent is really tough. A customer does not want to buy a screamingly loud machine from Dell, and to keep that from happening Dell has to spend a lot more time designing a solution that will stay quite and cool. I know for a fact that Dell did not exactly succeed with this. A friend of mine just received a Dell Pentium 4 3.6 GHz machine, and it whines very badly. Not only that, but under load the computer audibly spools up the fans to where they are quite annoying. This constant spinning up and down of fans is quite distracting, especially considering that my friend works mainly with audio, music, and sound effects for his job. I can only imagine what the 3.8 GHz version sounds like. When all of these factors are taken together, it makes no sense why AMD has not increased its marketshare by a significant amount since the introduction of the Athlon 64. Now that it is making its parts on 90 nm, they have a lot more production capabilities then when it was only producing Athlons at 130 nm. Not only that, but the latest Venice and San Diego cores run very efficiently. Add on top of that the recent introduction of their excellent dual core products, and one honestly has to question why AMD has not picked up more customers. While in the end AMD might not win the case, it will certainly allow them to take up more marketshare because of the intense scrutiny that Intel will be under about its business practices. While these practices might not necessarily be illegal, they are certainly unethical. With this scrutiny Intel will really have to toe the line in terms of marketing payments to manufacturers, as well as keep their product prices consistent from manufacturer to manufacturer. Without the threat of Intel withholding millions of marketing dollars because of the presence of AMD products, big companies like HP, Gateway, IBM, and others will more happily adopt AMD products. It is when tens of millions are held back by Intel for "non-compliance" that makes these companies nervous. When those millions could mean the difference of a small profit that quarter vs. missing its earnings, these companies will do a lot of things to keep their shareholders happy. We currently have a thread open on this, so please join in the discussion! Site News I am finally getting my grubby paws on a GeForce 7800 GTX. Leadtek is being exceptionally nice and shipping out a card as we speak. I couldn't be happier! A big thanks to Isaac and Leadtek for making this happen. I must admit, when I spent my time as a small system builder, I was very happy with Leadtek cards. They always seemed to have great production values, and the cards rarely had problems. While all of the 7800 GTX's out at this time are essentially made by NVIDIA and sold to its partners as a whole (vs. the partners getting the chips and spec'ing out the full fledged board), Leadtek is doing their best to keep their products as different as possible. Right now this includes a faster clockspeed (450 MHz core and 1250 MHz memory), VIVO functionality, an outstanding bundle, and a solid 18 month warranty. Should be in this week! My Trouble with ATI? article may not be 100% correct, but it looks as if the claims of ATI experiencing poor yields and leakage problems with the R520 are not unfounded. Digitimes is reporting that yields and leakage are problems for ATI, and this has been the reason why we haven't seen this product yet. I no longer think the "32 pipeline design and can only get 16 of them running" theory is not entirely correct though. My thoughts on this is that the R520 will feature 16 ROPs, but that its "shader pipeline" count will not really be applicable to other current offerings (like the 7800 GTX, 6800 series, and X800 series). I think the problem here is that ATI is trying to push its performance with high clockspeeds (possibly in the 600 to 700 MHz range). This is significantly higher than the 7800 GTX, which is clocked at 430 MHz. Both NVIDIA and ATI make extensive use of standard cells in their designs, and these cells are not nearly as efficient as a full custom design. Using standard cells make transistor placement and routing much easier than a full custom. While there are still custom parts of every design, the majority of any ASIC by these fabless IHV's are made up of standard cells. Pushing the speed up on a design comprised of standard cells is much harder than a full custom part, and to get the speed up, more power has to be used. This has the side effect of more heat produced, and at 90 nm there is a greater likelihood of increased leakage. Leakage can be avoided with a good design, but placing 300 million+ transistors comprised mainly of standard cells is no small task, even with the most advanced design tools that ATI and NVIDIA have at their disposal. Many things can be done to decrease leakage, but invariably it makes it a much more expensive part to produce (increasing the amount of space between transistors, using more exotic materials as the insulating layer, etc.). A good discussion of this is also taking place here. June 28, 2005 AMD Files Anti-Trust Complaint Against Intel - Josh What originally started in Japan, has now spread throughout the world. AMD has finally formally brought a complaint against Intel in the US for anti-competitive behavior. It is no secret that Intel pays much of the marketing costs for its "partners", and the amount of money given depends on what percentage of parts are Intel only. The list of partners that have knuckled under the pressure is actually quite long, and these companies include HP/Compaq, Gateway, Toshiba, Acer, Sony, Office Depot, Best Buy, Circuit City, and many others (apparently 38 in total). There is a fine line here between being a competitive business, and unlawfully maintaining a near monopoly. Intel has set the bar very high for any kind of competitors to get into the X86 market, and AMD has clung in there for many years. Now it seems that AMD has had enough of Intel's practices. It is one thing to offer rebates and marketing help, but it is another to withhold these contractual agreements if the company in question has dealings with AMD. Withholding products and money from these companies seems to be a common complaint. Many companies report that they are promised some form of retaliation if they use AMD chips at all. This type of strong-arm tactics is definitely illegal, but for many years it has been overlooked. The crazy thing about this is that Intel doesn't really need to pursue this type of business behavior. They have products that are competitive with AMD, and they have far more fabs producing processors than AMD does. AMD does not have the ability to address more than 25% of the market with their current manufacturing abilities. So why does Intel continue to do these things? That is a very good question. There is no denying that AMD has a compelling product with its Athlon 64 brand, and the mobile Athlon 64's and Turions should provide good competition to Intel in the mobile space, but because of the pressure that Intel puts on these companies, AMD's parts are not being offered nearly as much as they probably would be. A good example of this was Emachine's Athlon 64 lineup of notebooks. These notebooks sold very well, and were typically hard to find. Gateway bought up Emachines, and suddenly these SKU's dried up and went away. Currently there is only one Gateway/Emachines laptop that features the Athlon 64, and it is only available at Best Buy. Apparently marketing money and Centrino supplies were withheld, and Gateway discontinued a very successful line of laptops due to it. I will be very interested to see how the courts will rule on this case. With the Japanese courts firmly behind AMD, I am sure the US courts will follow after a lengthy investigation. AMD has built up a good 20 years worth of evidence against Intel. Looks like things might get shaken up a bit! If you don't believe me, take a look at both Best Buy and Circuit City and count how many Athlon 64/Sempron machines and notebooks are offered. (Here is a hint... there are less than 20 models between the two online stores). In Best Buy's case, there are 57 Intel based products, and 4 AMD products. Also, as Chris Tom has been reporting, many of the AMD computers are typically located in the back of the store, or in strange nooks and corners. Intel machines are usually always showcased. Big money talks, there is no doubt.
If you have found this article interesting or a great help, please donate to this site.
Copyright 1999-2005 PenStar Systems, LLC. |
|||
