   |
|||
|
|||
|
:
News
: Archives
|
|
|
|
|
|
|||
|
|
|||
|
|
|||
|
May 2, 2005 Spring in the Rockies? - Josh If you like snow, you would love it here. It has been snowing in Laramie since Wednesday, and the forecast hints at it doing so until this Thursday. We do need the moisture, but my wife is certainly unhappy about some of her plants outside in the garden. Poor Tina. I usually don't mind snow and cold weather, but my house has been a bear to heat all winter! I want to stop using so much natural gas! My checkbook wants me to stop using so much natural gas! As I am writing this, it is supposedly 18F outside. I sure wish Spring would get here. I had honestly expected NVIDIA to release its next gen product by now (well, I had originally expected this in early March, but the signs started to show differently soon after that). Apparently the G70 is up and running nicely at NVIDIA's headquarters, but I haven't heard how production is going (there have been conflicting reports as to how far along production is). My gut feeling is that they are actually fabbing production silicon, but finished product won't start rolling out until later this month. This probably means an early June introduction, probably meant to coincide with ATI's R520 release (which is supposed to be released during Computex). The G70 does appear to be based on TSMC's 110 nm process, and not 90 nm as some had surmised. It will be a very big chip though, and I imagine power and cooling are going to be two very large aspects for it. My SLI testing is about done, and I have been very happy with how it works. It is pretty seamless, and very smooth to run. Taking games to high resolutions with good levels of AA and AF applied has been a lot of fun. It honestly leaves me salivating for what the next gen of cards will bring to the table, as well as what those cards will act like in SLI! I can only imagine what 48 pixel pipelines and 10 vertex shaders running at 450 to 500 MHz would do for even the toughest games around. One thing I will be very curious to see is if AA will work with NVIDIA's HDR provided by its FP16 frame buffer. Currently in the NV4x architecture, HDR effects enabled by the FP16 frame buffer are incompatible with their current multi-sampling AA unit. I also have Chaintech's VNF4/Ultra motherboard on the bench, and it is a very solid little board. For the budget enthusiast, I can think of few other products that offer what it does for the extreme low price that they charge for it. The average price for this board is anywhere between $85 and $105 US. That is a lot of motherboard for not a whole lot of scratch. I also have an Athlon 64 3800+ based on the Venice core running in the SLI setup, and it has provided that extra bit of performance to really push the SLI cards. I had previously only had a 3200+, so this upgrade has been a Godsend for my testing. Thanks AMD! That is all for now, I will try to get my MSI K8N Neo 4 Platinum review up this week, but this evening you will be seeing a review from Scott (poor, long lost Scott). So keep reading! April 25, 2005 Microsoft Releases WinXP 64 Today - Josh In a move that AMD certainly has been waiting a long time for, Microsoft is releasing its Windoows products supporting X86-64. Windows XP Professional x64 Edition and Windows Server 2003 x64 Edition are the two products receiving the 64 bit treatment, and it is a welcome release for many consumers. AMD packaged a RTM (release to manufacturing) version of WinXP Pro 64 with a 3800+ Venice core processor, and I had the chance to test it out. Other than running into a few snags with the installation (the ATI 64 bit drivers failed to completely install the first time) I thought the 64 bit version ran very smooth. 3D games, benchmarks, and other operations ran very nicely. The ability to run 32 bit applications on a 64 bit operating system was excellent, and everything that I threw at it ran great. There is one looming issue that cannot be overlooked though. Driver support. While companies like NVIDIA, ATI, Realtek, and others have beta 64 bit drivers out, the majority of products out there have no support. This of course could change radically with the official release, but I am not holding my breath. Creative is one company that has not always had the shiniest driver support. Currently only the Audigy cards have any kind of 64 bit support, but the rest of their products are sorely lacking. For those hoping to have drivers for their MSI K8N Neo 4 Platinum/SLI board with its built in 24 bit Live! chip, they may have to look elsewhere for the time being. Another area that looks like it is going to be hammered are telephone modems. Doing a search around the web has revealed very few telephone modems having 64 bit support. For users wanting to get WinXP 64 and don't have a broadband connection, then internet access could be a very tricky situation. Again, I hope this changes with today's official release, but I still am not holding my breath. There are two big features of the 64 bit release and the first is the increased memory addressing without the use of PAE (physical address extensions). 16 terabytes are now addressable (theoretically) by a single processor/machine, but for the time being I believe we will see around 8 to 16 GB of addressable memory as the maximum per processor. The second feature, and probably the more useful for most users, is the newly exposed registers that X86-64 brings to the table. X86 has been plagued by the overall lack of registers that it officially supports, but CPU makers have gotten around this by doing tricks like register renaming to give more actual registers than what is officially supported. X86-64 exposes 4X the amount of registers that the standard X86 allows. This is what probably gives the Athlon 64 that 5% to 15% performance increase over the same processor in a 32 bit environment. This allows the core to be much more efficient in its data usage, and it keeps the pipelines more adequately fed with data. Initial reports show WinXP 64 to be a very solid OS. While there are plenty of changes under the hood, the end user should notice very little between the 32 bit OS and the new 64 bit OS. Again, the software support is seamless, and the WoW (Windows on Windows- which allows the use of 32 bit code on the 64 bit OS) performance is actually quite incredible, especially in light that it is a "compatibility mode". AMD certainly wishes that Microsoft had released this OS a year ago, before EMT64 was officially exposed by Intel. But all things being equal, the industry really wasn't ready to support a mainstream 64 bit desktop OS at that time. Looking at the current driver situation, I think it is safe to say that the support coming out now is much more robust than what would have been available a year ago. This has also allowed Microsoft a lot of time to iron out the bugs and secure the OS against any vulnerabilities. While there will still be bugs and vulnerabilities that will be found, having an extra year to work on these things certainly makes for a better OS. April 22, 2005 Thoughts on SLI Interview and Technology - Josh So far the SLI interview has been a pretty good success for the site, and I have received some good suggestions as well as feedback from the interview. I would like to thank Chris Daniel and Brian Burke for working with me to make this interview happen. I personally thought there was a lot of good information in there, and Chris did an excellent job of conveying many of the more complex technical aspects into language most of us can understand. One thing that I was very impressed by was the professionalism that Chris exhibited, as there was no trash talking or competitor bashing. It was just a good, solid overview of the SLI technology, the advantages it brings to the table, and a truthful look at some of the possible disadvantages. SLI is still a very new technology, and there are hurdles that NVIDIA still has to overcome, but for those willing to spend the money on a SLI rig, it can be a perfect fit for their needs and wants. A couple of things came up about the interview and the technology, and I thought I would pass some of it on. One of the more interesting questions that was brought up is the ability for cards like the Gigabyte GV-3D1 (dual 6600 GT on a single card) to possibly be used in SLI? Currently the GV-3D1 does not have a SLI connector, but is the SLI design flexible enough to handle such a product in the future? One question that I had is what communications protocol is used between the two video cards? Is it some proprietary link? Or is the technology based on HyperTransport? Since NVIDIA is a leading member of the HyperTransport consortium, one would think that they could possibly use the high bandwidth/low latency HT protocol to effectively communicate between the cards (especially considering that HT is bidirectional, and might fit hand in glove to such usage). One other area that I have thought long and hard about was that of the apparent driver overhead when SLI is enabled. Chris mentions in his interview that there is no appreciable driver overhead for the AFR and SFR rendering solutions. If this was true, then why does a SLI setup run slower at lower resolutions than a single card? After thinking about it, as well as receiving some hints on a couple of forums online, it seems that the communications and synchronization between the two cards is causing this to happen. The over the top connector gives 1 GB/sec of bandwidth, but when each card is trying to render a frame and then transfer the frame content information to the other card as well as make sure everything is synchronized, that turns out to be a massive amount of data to transfer. It becomes clear that the bottleneck is the connector in such situations. Now, this situation really only applies to low resolution rendering without AA or AF, and frame rates are sky high already. Of course, where SLI really shines is at high resolution with high quality settings enabled. When frame rates are hitting between 50 and 200, then the connector seems to do fine. Only when rates get above that do we see the bottleneck. Running Doom 3 at 1600x1200 at 4X AA and 8X AF at a solid 50 to 60 fps is nothing to sneer at! That is all for today, so have a great weekend! Expect a case review from Scott on Sunday though! April 21, 2005 AMD Releases Dual Core Products - Josh Today is the day that AMD is finally officially releasing their dual core products. Last summer AMD made the announcement that they had finished taping off the first dual core design, and by August they already had working samples. AMD has also been providing its partners with dual core chips since January, and so there are actually products available for sale using these chips. AMD further threw a wrench into the works by announcing its Athlon X2 products, which are dual core chips for the desktop and workstation market. These will not officially be available until June, but AMD probably thought to counter Intel's recent dual core desktop announcement with this one. In four more days AMD will get another boost, and this time from Microsoft. Windows XP Professional 64 will be officially released to much fanfare, and the operating system that many Athlon 64 users have been waiting for will finally be delivered. I am unsure of the driver support at this time, but many of the major component manufacturers have released 64 bit drivers of their products. There are still quite a few though that haven't worked terribly hard on this release (as of yet). The maximum speed of the dual core products will be 2.2 GHz, and in the future this number will rise. What is impressive about this processor is that even though it has two full cores, it only consumes about as much power as a 2.6 GHz FX-55. AMD has tweaked the design and process for this product so that it consumes significantly less power. The interface between the two chips is a lot more engineered than Intel's dual core processors, and the cores intelligently share resources such as the Hypertransport links and the dual channel memory controller. These chips also use AMD's MOESI cache coherency protocol, and the cache exchanges between the two processors is done at full processor speed, so it happens very quickly with a lot of bandwidth and low latency. This is a huge event for AMD, as they beat Intel to the dual core Server market. This is the one area where a dual core product is greatly desired, much more so than the desktop space. The server and database guys running quad Opterons simply cannot wait to get their hands on the dual core chips so they can utilize 8 total processors without having to upgrade their motherboard/system. This is also the two year anniversary for the Opteron, and its acceptance into the industry has been impressive up to date. AMD being able to supply dual core products to the server market will help it compete in this very lucrative market. While the dual core chips will help the bottom line somewhat, it really won't be until Q3 that the dual core strategy will really start to make an impact on AMD's financials. By that time AMD will be busy selling dual core Opterons, and the dual core desktop chips will start to make an impact. Supermicro finally threw its hat into the Opteron market, which is quite a coup considering that Supermicro has NEVER support an AMD product (officially that is). Now only Dell stands alone. Even they might be feeling the pressure, especially from their server customers. With Intel not scheduled to release a dual core Xeon anytime soon, I think that Dell has to seriously consider the Opteron. So what does Dell do? Keep its primary partner (Intel) happy? Or make their customers and end users happy? That's a tough one! The Tech Report has an awesome review of the latest AMD dual core processors (though no X2's). Scott goes over quite a few benchmarks, and the differences in performance can be staggering at times (depending on how multi-threaded aware an application is). In some apps, the dual core 875 processors just show a staggering lead over everything else. Highly recommended that you read this! Anand also lets fly with its review. News of Interest - Josh BFG has announced that they are shipping their 6800 Ultra OC 512 MB boards to Falcon Northwest. Initial availability of computers using this part should be late April, with retail penetration by the beginning of May. These boards use 16 GDDR-3 chips, which required a PCB respin from the standard Ultra board design. So far there have not been GDDR-3 chips of high enough density to enable the use of 8 memory chips. The BFG 6800 Ultra OC runs at 425 MHz as compared to 400 MHz for the standard Ultra. The memory is not overclocked. This is one serious board, and I bet it can be had for a very serious price! The nice and friendly folks over at the [H] have one of the first reviews of the Radeon X850 XT-PE AGP video card. This card is based on the R481 AGP native chip from ATI. No Rialto on this puppy! So, for those looking for the fastest 3D card in the world, yet still keep their AGP motherboard, this review is for you! Now, when we will see these cards in retail... Hot Hardware has a couple of reviews that I passed over recently. The first is a good look at a pair of Asus 6600 GT's in SLI. They check out and see what kind of performance advantages a user can get with two of these midrange cards. They also look at the EPoX 5LWA+ (where do they get these numbers and letters from?) which is based on the i925XE chipset. April 19, 2005 NVIDIA SLI Interview - Josh I was recently given the chance to compile a bunch of questions regarding NVIDIA's SLI implementation. Chris Daniel, the product manager of NVIDIA's SLI program, was kind enough to answer these questions in a very thorough manner. There is information presented here that I have yet to see anywhere else. A very good read if you are interested in SLI! Here is a quote:
You can read the entire article here. AMD and Their Process Technology - Josh Lately I have been very impressed by what AMD has done with their process technology. Through cooperative efforts with IBM, Chartered, and others AMD has really maximized their potential when it comes to utilizing advanced process technology on a budget basis. AMD does not have the billions of dollars that Intel has to develop their own process (not to mention all the Fabs that Intel owns), and so has to rely on partnerships to create a process that is competitive with its much larger counterpart. AMD realized very early on that it could not go head on against Intel by itself, and for the past 15 years AMD has been quietly acquiring partnerships with some very high profile companies. These include Motorola, IBM, Infineon, and others. Since these companies are not usually direct competitors for AMD, it is easy for them to enter into cross licensing agreements when it comes to process technologies and transistor design. If we take a look at AMD's past, we can see how far they have come in terms of their process implementations. At the 350 nm node, AMD had a solid process. They were able to produce the K5 and K6 processors fairly effectively, but with just the Austin fab and the test fab in California, AMD was never able to produce as many chips as they needed to. 350 nm was a clean process for AMD, and the K6 chip took to it fairly effectively. Things were not so shiny once AMD tried to go to 250 nm. This was a tough, tough node to get to for AMD. While AMD had pressured Intel with the K6 233 MHz chip, they were unable to jump to the next speed grade without the transition to 250 nm. Intel didn't have that problem. Intel was able to fend off all advances from AMD with the 333 MHz Pentium 2, and then the 350 and 400 MHz versions. AMD was stuck at 233 MHz for a long time. AMD did a tremendous amount of work to get 250 nm up and running effectively, and they were 6+ months behind in their processor release schedule. Eventually the engineers got the process up and running as it should, but the K6 was a very difficult design in many aspects to get up to speed. AMD's 250 nm process, with all its tweaks, turned into a very clean process. The only real issue at that point was the design of the K6 not wanting to clock to high speeds. When the Athlon was first produced on AMD's 250 nm process, it took off like a rocket. AMD was initially going to release a 450 MHz version at first, but they found that most of the dice ran at 550 MHz+! 450 MHz was never released, and the lowest clocked Athlon was 500 MHz. During this time AMD was able to one up Intel at every clock speed increase, mainly due to the excellent overall design of the Athlon and the rock solid 250 nm process that AMD had worked so hard at. Eventually the Athlon hit 650 MHz (and if I remember correctly, there were some 250 nm 700 MHz Athlons). The 180 nm process was one of the first to feature copper interconnects. Working closely with Motorola, AMD created a fantastic 180 nm process that took the Athlon to 1 GHz and beyond. Intel had a non-copper 180 nm process, but it did feature some very interesting aspects that did lend itself to fast transistor performance. Unfortunately for Intel, the Pentium !!! design was not as scalable as the Athlon was. Intel barely lost the race to 1 GHz, but when the actual numbers of 1 GHz chips shipped during the 6 months after 1 GHz was reached, AMD was shipping around 90% of all 1 GHz products (and Dell was getting the vast amount of 1 GHz P!!!s). Intel then swung into action with the Pentium 4, and clockspeeds were not an issue for them anymore. AMD's 180 nm process took the later generation of Athlon XPs to 1733 MHz. In hindsight, it was a very successful and flexible process for AMD. 130 nm was a tough one for AMD. First off they were approximately 6 months behind Intel in terms of shipping 130 nm parts to the market, and the second issue was that their 130 nm bulk process was not very fast. The first Athlon XP to be made on the new 130 nm process was slower than expected, and it put off a significant amount of heat. The T-bred A was clocked at a maximum of 1.8 GHz, and even with extreme overclocking these chips could not reliably reach 1.866 GHz all the time. While AMD was able to produce more chips due to the smaller die size of the Athlon, it wasn't until the T-bred B was released that AMD was able to clock their speeds up to what they should be. Not only was the core redesigned, but an extra metal layer was added to the process. These two things helped AMD immensely, and the clock speeds again shot up to where they needed to be. 130 nm was a solid process for AMD, and it allowed them to stay fairly competitive with Intel, but not without some really nasty teething pains. 130 nm SOI was a huge gamble for AMD. This was a technology that was very new, and very hard to implement. Working extensively with IBM, AMD was able to integrate SOI into their 130 nm process, and the ramp up was significantly faster than when AMD moved to the bulk 130 nm process. The Athlon XP would never receive the benefits of SOI, but rather AMD would exclusively utilize this technology for the then upcoming Athlon 64 and Opteron products. SOI has several different effects on a processor, and most of it is focused on making the transistors less leaky, as well as accelerating overall transistor performance by about 20% to 30%. The initial Athlon 64's were clocked at a modest 2.0 GHz (Athlon 64 3200+), but quickly jumped to 2.2 GHz and then to 2.4 GHz. The fastest clocked Athlon 64 on the 130 nm Low-K process (with some strained silicon thrown in for good measure) is the FX-55 which is clocked at 2.6 GHz. This is again a very impressive gain due to the outstanding design of the Athlon 64, as well as AMD's world class 130 nm SOI process. The jump to the initial 130 nm process was a huge one, and it made AMD's life miserable for a while. Many experts around the industry saw the jump to 90 nm as being almost twice as hard as the one to 130 nm. This did not seem to be the case for AMD though. Due to its strength with their 130 nm SOI process, as well as the Athlon 64's design, AMD was able to compete well against Intel and their 90 nm Prescott products. AMD took their time with 90 nm, and was very methodical in their implementation. While AMD was about 9 months behind Intel in terms of 90 nm production, their 90 nm products did not exhibit the power and heat issues that Intel struggled with. This is in part due to the use of SOI, as well as the overall transistor design. Intel is pushing clockspeed with their products, and so the transistor design they use has to switch very fast, and it requires a lot of power. As power is increased, so does leakage. By using a slower transistor design and requiring a lower clockspeed to keep up in terms of performance along with the advantages SOI brings to the table, the Athlon 64 is a cooler and more power efficient processor than the current Pentium 4. AMD has truly done wonders with their process technology through the years, and the 90 nm culmination of this really shows how talented the AMD engineers are. This is not to say that Intel does not have talented engineers, but when we consider the budgets the each company has to work with, AMD has done a fantastic job on a virtual shoestring (though I wish I could live on that shoestring!). AMD is also well on their way to begin testing of its 65 nm process that will also feature 300 mm wafers in the new Fab 36 (right next door to Fab 30 in Dresden) in mid 2005. Once production begins in mid 2006, AMD will begin converting Fab 30 to 65 nm as well. Many estimate that AMD will have excess capacity on hand, and may be willing to rent out Fab space. I am sure companies like NVIDIA, ATI, VIA, and others are very interested in possibly utilizing that space. Especially if AMD is as successful at 65 nm as they were with the current 90 nm process. I guess only time will tell! April 18, 2005 Monday Morning Musings - Josh This weekend was spent babysitting the kid, and trying to wrap my head around how to do the SLI review. Things were complicated even more by my acquisition of a Rev. E Athlon 64 3800+ from AMD. Luckily the MSI SLI motherboard I have (the K8N Neo Diamond) runs the Rev. E without a problem with the 3.10 BIOS. I am still not sure why I received the Diamond vs. the regular Platinum/SLI since I am in the US. Strange, yet odd. So far this motherboard I have does not want to overclock worth beans, so the 3800+ review will have to wait for another motherboard (which luckily I have). NVIDIA has certainly been pumping out the driver releases as of late, last month brought us the 71.84's and this month has given us the 71.89. Not a huge jump in numbers or performance, but it is certainly a step forward and it does add some features as well as more SLI profiles. I hope that this is a sign that NVIDIA will try to match ATI's Catalyst program with official driver releases. Customers and OEMs like good, solid drivers that fix problems with applications when they pop up (and new games are always being released, and typically have teething pains at the beginning with the latest crop of drivers). While we may not see a monthly release from NVIDIA, I think we will see much more driver attention being paid to users (especially since SLI really is picking up, and users will demand more built in profiles for games). If you by chance read my latest State of 3D, then you probably remember me saying that perhaps we have seen the first G70 chip from NVIDIA in the form of the GeForce Go 6800 Ultra. Well, that is not the case, but I think it is a good indication of things to come. The Go 6800 Ultra is in fact the NV42M, the mobile version of the NV42 which is made on TSMC's 110 nm process. This is a true 12 pixel pipe/5 vertex shader design (so no unlocking of extra pipelines for the modders out there) with the fixed video unit. It may have a few pipeline optimizations, but nothing that really makes it stand out against previous NV40 versions. What is impressive is that this large of chip runs at 450 MHz on TSMC's 110 nm process. When the NV40 was initially released, it had a heck of a time clocking anywhere over 415 MHz, much less up to 450 MHz! Now we are seeing these parts go to 450 MHz in a mobile application (where cooling and power are two very large components to a successful design). TSMC has really done wonders with its 110 nm process, and I feel that it offers about 90% of the overall transistor performance of the much more expensive 130 nm Low-K. I still believe that NVIDIA will release its next gen product on 110 nm, though it is starting to look like an early summer release vs. the late April that I initially suggested. NVIDIA is keeping the heat on ATI, and the upcoming release of the 512 MB 6800 Ultras should keep that pot at a boil until we see the next gen of products. The very fact that NVIDIA is still releasing an upgraded top of the line 6800 Ultra shows me that the next gen of products is not quite here yet. Is it that the chips and boards are not ready? Do they want to build up stock so that when they release the card there will be a solid supply of products? Are they simply waiting for ATI to release their R520 and are hoping to foil that release? Many questions, and very few answers are coming out. NVIDIA has certainly done some good things with their NV40 chips from IBM. The first several months of production were REALLY rocky, and there were jokes made about "how many wafers per good die" from pundits around the industry. Since about July their issues were mainly solved, and right now their production of parts is very, very good. Not only that, but often the latest Ultra binned chips can reach 450 MHz and above. I know that the two 6800 GT boards in SLI are running at a very respectable 415 MHz, and for both of them to reach that speed while in SLI (and only a 480 watt power supply) is something to respect. So, looking back, things have been pretty rough for NVIDIA in terms of getting product to market at the high end, but as the way things are sitting now, it all worked out ok. Undoubtedly their biggest success story for the past year was the release of the 6600 GT in both PCI-E and AGP form. This has been a very good seller for NVIDIA, and it has certainly bolstered their bottom line. Their other success story was NVIDIA's ability to recapture a large portion of the retail market. While ATI focused on the OEM space and was the dominant player there throughout last year, NVIDIA was able to satisfy their OEM market and pushed a lot of product out of retail. As you recall, ATI had some real yield issues with their X800 series of chips throughout most of last year. Right now both companies are able to adequately supply the market with their high end chips. The X850 XT and XT PE cards are certainly excellent performers, and are essentially the fastest cards on the planet (not counting SLI 6800 Ultras of course). News of Interest - Josh Speaking of the Go 6800 Ultra, the [H] takes a look at the Dell Inspiron XPS Gen 2 notebook, which takes laptop gaming to an entirely new level. Very good reading, and a very impressive performer (to say the least). Gaming with new titles at 1920 x 1200 sure sounds good! Wish Dell would somehow stumble across my site and offer up a laptop for review! Anand takes a good, hard look at the latest AMD Sempron 3300+. This is the Rev. E variant of the Sempron, with all the extra goodies that the revision brings. 90 nm and some overclocking potential come for the low price of around $120 per chip. While the Socket 754 is now relegated to the budget sector, I am still using a 754 for myself with a 3700+, and I have no complaints about overall performance! Supposedly the official Intel dual core launch is today, but I am surprised that the NDA has not passed as of yet. So, keep tuned, and we will see what shakes out! Another thing to watch out for is an exclusive SLI interview that I conducted last week. I am scheduled to get everything back today so that I can hopefully post it. Some good info in there, and I was able to ask a few questions that I had not seen the answer to anywhere else. April 16, 2005 MSI Makes the Necessary Changes - Josh Color me impressed. The day after NVIDIA contacted MSI about the 6800 GT/GTO situation, MSI has started to change their marketing materials for this product. The first change we see is on their website showing the 6800 GTO product. The product page used to say it was a 6800 GT, and it had the official NVIDIA 6800 GT logo on it. As you can now see, MSI is now calling this part by its correct name. MSI really can't do a whole lot with the products that are already out in the market, and I wonder when they will finally get a new box that advertises it as a GTO product. Since they print these boxes up in the thousands, I am guessing it will be a while. Oh well. Still, I am quite happy to see MSI correct this mistake, and do right by the consumer. Strong work to everyone involved in making this right! NVIDIA Releasing More Drivers - Josh ATI has really set the standard with their Catalyst Crew program, and while their drivers may not always be a huge jump from month to month, the very fact that they do release a new, stable, WHQL certified driver consistently is something to be very impressed about. Most reviewers comment directly on this, about how nice it is to have a new driver for their product every month, and I know OEM's certainly appreciate it. In previous conference calls we have heard mention of NVIDIA hoping to do one or two WHQL releases a year, but I think that they are feeling the heat from the industry about this stance. Last month they released their 71.84 drivers, and just a couple of days ago they released their WHQL candidate 71.89. Not a huge jump, I know, but it does have some improvements and better SLI support. I am quite happy to see NVIDIA updating their official drivers more regularly now, especially since there are so many new gaming titles, not to mention uncovering new functionality in their products (eg. better support for PureVideo, etc.). The competition from ATI and NVIDIA has resulted in essentially a quantum leap in rendering performance in the past couple of years. The $200 price point is very competitive, and though the 6600 GT is the faster of the products here, ATI is also competing on price (eg. the X700 Pro's are cheaper or have more memory than the 6600 GT's). In either case, the user will get a very good gaming experience at that price. The X850 XT PE is the fastest gaming card around, but the 6800 Ultra still provides competition. When you throw SLI into the mix, things get even more interesting. I currently have a 6800 GT SLI setup I am testing, and so far I have been very impressed with how well it works. It is also quite easy to create SLI profiles for games that are not officially supported. Running Morrowind at 1024x768 at 16X AA (this setting is uncovered in Rivatuner and n-Hancer) is simply breathtaking, and surprisingly smooth. 16 jittered AA samples per frame makes for a very nice scene. ATI will be releasing their multi-card technology sometime in late summer, so once that hits we will see how the differing philosophies of the two companies compare. Not only do I have the SLI setup I am working on, but I have a pretty comprehensive interview that I will be posting either Monday or Tuesday with NVIDIA about their SLI technology (I asked some questions I have never seen addressed anywhere else). Add to that AMD just sent me their Athlon 64 3800+ Venice chip, along with a RTM copy of WinXP 64. I guess I have my week set out for me! New Archives Section - Josh I have been saving up my previous front page news in an archive form, but I never really got around to organizing it so people could check out some previous news posts. Well, now it is up and running. If you are curious about what I thought was going on in 2003, this is the place to look! Enjoy the weekend! April 15, 2005 Official Response from NVIDIA on MSI and the 6800 GT - Josh Late yesterday I received a response (very prompt I might add) from NVIDIA about the MSI 6800GT/6800 issue I had pointed out here. Brian Burke was kind enough to quickly respond to me with this timely quote:
I have not heard anything from MSI about this as of yet, but I am curious if they will take the "suggestions" from NVIDIA seriously. Honestly, I could care less about what a company calls a video card or motherboard, but branding a product as a higher end product, and then trying to fool consumers with this misleading branding is just terrible. While MSI is obviously not charging the same amount for this product as their true GT, it is still not a good practice. Hopefully we will see some action on this, but I don't expect much soon. News of Interest - Josh It seems that AMD slipped a couple of the new Sempron 3300's to a couple of sites. This product is based on the new Palermo core, which is a Rev. E of the Athlon 64 design. This has 128 K of L2 cache, has the 64 bit functionality disabled, features SSE3 and an improved memory controller, and is produced on AMD's 90 nm process. Hot Hardware features a review on this number (Jeff did a very nice writeup), and AMDZone also did a quick once-over with this chip, but due to some motherboard issues was not able to complete it. This appears to be a very solid, and exciting, budget processor for the Socket 754 crowd. This chips showed decent overclocking ability, and I imagine with a little work that they will hit 2.5 GHz with few problems. AMD really has an excellent budget processor, and now that we are starting to see PCI-E Socket 754 motherboards, I would imagine that many OEM's are looking forward to these CPUs hitting the market. [H]ardOCP is taking the big step into doing full system reviews. Before they jump in those waters though, they are explaining their philosophy behind these tests, and how they will perform them. I certainly wish I could do a fully system review! Unfortunately, not many OEM's are terribly keen on sending me a $2800 system for testing. Anandtech takes the new NVIDIA nForce 4 Intel Edition and does a direct comparison to the upcoming Intel i955X chipset. This is a very good overview of the performance that NVIDIA can bring to the Intel platform, and considering this is NVIDIA's first Intel chipset (not counting X-Box) I think they did a tremendous job with their design. The nForce 4 IE matches, and often beats, the i955X product in a variety of situations and benchmarks. Add to that the ability to utilize SLI, and NVIDIA has a winning product on their hands. Of course, the nForce 4 IE is not an inexpensive solution, and will be relegated to the Ultra High End market for Intel based motherboards. Expect to see these boards retail for above $200 initially. In a review that I had overlooked some weeks ago, Ratchet from Rage3D (a really good guy who also happens to write really well) takes a look at the HIS Radeon offerings (X850XT and X800XL). These puppies come with a very non-standard ATI heatsink, and have some very nice features and bundles. A very good read for those interested in these cards. OCTools is giving away 4 GB of Corsair Xpert memory. This is the stuff that has the really nifty LED readouts on them, and is pretty solid memory to boot. Sign up here. Corsair (one of my sponsors... for the moment) has been making some rather large waves as of late. They are working with many Tier one motherboard manufacturers to enable the best possible memory performance out of NVIDIA's nForce 4 Intel Edition. Their XMS-2 modules are clocking very high, with very low latencies (for DDR-2 that is). Very solid stuff, read more about it here. Neoseeker has the very inexpensive, but impressive, Chaintech VNF4/Ultra on the testbench today. This is a very budget oriented motherboard, but it is still based on the nForce 4 Ultra chipset. While it may not have all the bells and whistles more expensive nForce 4 Ultra boards has, it certainly keeps up in terms of performance and overclocking. A very good read. In other news, I finally got back into brewing last night. I had forgotten how much I enjoyed brewing in the past, and it was very fun and relaxing to get down and brew some beer. I know it is the wrong time of the year for it, but I couldn't resist brewing up an Oktoberfest! Sure, sure, make fun of me now for doing it in the Spring instead of early Fall, but I haven't had a good Oktoberfest in ages and I deserve one (or 5 gallons... your pick)! I love the way my house smells after brewing, but I am worried that I may have added too many hops. I put in less than what the recipe asked for, but I still think it may be too much. Oh well. Hopefully it will mellow out and I will have a nice product in the end. I figure about a week and a half in the primary fermenter, then rack it for clarity for another couple of days. Add the primer, and I should be bottling in two weeks or so. Allow a minimum of 6 days to carbonate, and I can have my first bottle. I can't wait (though it really should be bottled for at least 21 days so as to smooth out some). Enough for now, have a great weekend! April 14, 2005 Bittersweet Results for AMD - Josh Late yesterday AMD announced its quarterly earnings for Q1 of this year. They had approximately $1.227 billion in sales with an operating loss of $46 million. Not exceptionally good news, but then again not bad news. I remember back several years ago when they almost broke even with revenues of around $750 million, so they are definitely heading in the right direction. The processor group reported record sales, and posted a very healthy $92 million profit in operating income. Sales of the Opteron/Athlon 64/Sempron products are all incredibly healthy, though AMD did not publish their average CPU selling price. Looking at the Athlon 64 and Opteron products, they are all retailing at $130 and above. Sempron products of course are typically lower than $120. The area of weakness for AMD is its Spansion memory branch that it runs jointly with Fujitsu. The same day that AMD released its financials was the same day that Spansion filed to sell common stock. This basically means that both AMD and Fujitsu are trying to divest themselves of Spansion, though each will remain major shareholders in the company. This will help the overall financials for AMD, but it also cuts off one of their legs. If we remember some years back when the AMD Athlon XP was not scaling in speed next to the Pentium 4, AMD suffered some very poor revenues from their processor group. Luckily for AMD, their flash memory division was having a heyday, and it helped to keep the company afloat during hard times. Now the cycle has come around where the processor group is doing very well, but the flash memory group is doing poorly. I have serious doubts as to why AMD would want to spin off Spansion as its own entity, but perhaps their position allows them a clearer picture of the future flash market. My opinion is that they see rough times ahead, and that the flash market will become pretty unprofitable to any company moving into it. So, making Spansion stand on its own feet could be a very good idea. But what if AMD's processor group falls on hard times again? What if several other Flash memory producers back out of the market and Spansion becomes profitable? AMD and Fujitsu appear to be in a better position to predict that situation than I am obviously. Still, AMD is primarily known as a CPU company, and its overwhelming success with the Athlon 64 and Opteron products have certainly put it at the forefront of the industry. Looking at both Intel and AMD, it just seems like the decisions AMD made in 1995 are really starting to pay off for them. Back at that point they bought NexGen and adopted their design for the K6. At that same time they started to bring in a lot of DEC guys and started to lay the foundation for the K7 series. This foundation would also be the basis for the K8 series, and the design decisions they made in terms of both architecture and process have proven to be pretty far sighted. While AMD did suffer in 2002/2003 due to the Athlon XP not staying competitive with the Pentium 4, the release of the Athlon 64 in late 2003 was the crowning and defining moment for AMD. Here was an architecture that could scale for the future, had many forward looking features that would help to eliminate bottlenecks, yet still provide excellent connectivity between the processor and host system, as well as other processors. HyperTransport, integrated memory controller, large caches, and their 64 bit functionality and performance all work together to give AMD a very profitable and flexible processor. While AMD has not gained huge amounts of marketshare, they do have a much more profitable product. Athlon FX-55's are selling for well more than $800, while the next step down, the Athlon 64 4000+, is selling above $500. Never in the history of AMD have they had desktop products that have kept their prices up this high for so long. Add to that the increasing popularity of the Opteron processor, and AMD looks to have some very exceptional products for the next year and a half. So what happens in a year and a half? It looks as if Intel will be releasing new products that will have many of the features of the Athlon 64. Intel's HTT technology (CSI) will be introduced to the server market and high end desktop, and it will also embrace an integrated memory controller. Now, Intel has a very impressive army of engineers and if they decide to push something, there is no holding them back. We can look at the upcoming dual core launch on the 18th of this month for support of that opinion. Intel should not be in a position to launch this product, but they have been working overtime to get this design out the door before AMD releases its dual core Opterons on the 21st. Still, Intel's dual core launch does smack of desperation, especially since AMD has been showing off working dual core silicon (and a highly integrated dual core at that) since August of last year and has been shipping CPU's to partners since January. Intel's dual core architecture is not nearly as integrated AMD's, but this shouldn't prove to be a big bottleneck for the product. Essentially Intel has take two Prescott Pentiums, put them on the same piece of silicon, and designed a "new front side bus" between the two that will pass back and forth the data needed, as well as connect the product to the "real front side bus" of the system. This is a brute force type approach that allows Intel to use the current Prescott design without making major overhauls to it, but with the link running at 3.2 GHz it will still be effective. AMD's dual core offerings share the memory controller as well as having the HTT connections between them. The K8 was designed at the onset to be integrated into a dual core package, and so architectural decisions made five years ago reflect that functionality. AMD is going through some significant changes right now, and with their total conversion to 90 nm by the end of Q2, they can expect to see greater numbers of processors produced, and this in turn will help to win more customers due to supply constraint fears being minimized. Also, the new 90 nm designs will scale better in performance, as well as add features such as SSE3 and the ability to address more memory at DDR 400 speeds at the desktop and server level. The processor group at AMD looks to have quite a few months of smooth sailing without many worries from Intel. This will of course not last forever, as Intel is a very aggressive company with some excellent engineers. AMD will need to continue to innovate and push their architecture, and it appears as though 2006 will be a time of great change for AMD as well. DDR-3 will not be available in quantity in 2006, so it appears as though AMD will adopt DDR-2 in that year. Socket 1207 will be the new architecture for AMD that will utilize DDR-2. AMD has certainly taken its time adopting DDR-2, but that is not necessarily a bad thing. AMD has shown that the Athlon 64 processor works great with DDR-1, and it scales in performance very well using just PC3200 memory. AMD has stated many times that it does not see the advantages of DDR-2 until that memory hits DDR-2 667 and DDR-2 800 speeds. So while Intel is busily pushing DDR-2, AMD will not cash in on that technology until it is very mature. Again, AMD's performance with just using DDR is still class leading, so there is no reason for them to try to rush DDR-2 support. So, with the spinoff of Spansion, AMD's profits should increase, which will elicit a sigh of relief from investors. Sure, AMD's total revenues per quarter will go down, but considering the health of AMD's processor group we can expect to see many more positive quarters ahead. Again, the only real issue that many have with this move is if and when AMD's processors are not as competitive with Intel's, then AMD will not have another product to fall back on to keep revenues up. AMD needs to keep making good decisions, and consistently be more forward looking than Intel, or else Intel will come back and bite them where it hurts. AMD has a great thing going, and they need to capitalize on every mistake Intel makes. If they fail to do so, then 2007 could be the year Intel comes screaming back to processor dominance. Discuss this here. April 13, 2005 When is a 6800 GT not a GT? - Josh Back in February X-Bit wrote that MSI was shipping NV41/42 (which make up the GeForce 6800 cards) in 6800 GT boxes. This caused quite a confusion, but the situation was not really covered nor followed up on. Just recently I was perusing the video card stocks at a distributor that I use, and I noticed a MSI GeForce 6800 GT PCI-E that was selling for around $345. I thought this was a great price considering that most of the PCI-E versions of this card were over $390 and above. Then I started to look into the situation... Sure enough, the 6800 GT that MSI has listed is not actually a 6800 GT. It seems that MSI feels the need to call its 6800 offering a 6800 GT, and then they differentiate the SKU's by one letter: the real 6800 GT has the product code of NX6800GT-T2D256E, while the "12 pipe" 6800 GT is NX6800GTX-T2D256E. If you notice on MSI's product page, there is also the true 6800 listed, but it has a separate SKU (the NX6800-TD256E). I think this is extremely misleading! Apparently MSI feels they can label the GTX product as a GeForce 6800 GT because it is clocked at 350 MHz core. Buyers who think they are getting a true 6800 GT are getting short changed. Sure, they may pay a little bit less, but when NVIDIA listed the chip as a GeForce 6800 GT, it is supposed to be a NV40/45 with 16 pixel pipelines clocked at 350 MHz. MSI's product is a GeForce 6800 clocked at 350 MHz (just like Chaintech's overclocked 6800), and I hardly think it proper to call it a 6800 GT! I am trying to get some comments from MSI and NVIDIA, but so far nothing has been said. This is an unfortunate situation, especially for consumers. At least on MSI's website they do list that the product only has 12 pixel pipelines, but it is in small print and not immediately obvious. So, when you think you are getting a good deal on a PCI-E MSI GeForce 6800 GT, be sure to really look at what you are buying! April 12, 2005 Albatron GeForce TC 6200Q Review - Josh A little while back the nice folks at Albatron shipped their latest budget entry into the video card market. Since a good portion of all PCI-E enabled computers are being shipped with the Intell 900 series integrated graphics, there is a large potential market for a competent yet inexpensive PCI-E video card. NVIDIA hopes to address this market with the GeForce 6200 Turbo Cache, which is what the TC 6200Q is based on. This little number turned in some pretty impressive performances throughout a variety of benchmarks and realworld gaming experiences. Here is a quote:
You can read the entire article here. April 8, 2005 Random Thoughts on PCI-Express - Josh I love working with computers. It is as simple as that. My first computer was a Compu-Add 8088 knockoff based on the NEC V710 chip (running at a blistering 10 MHz), and while I was poor for a couple of years and couldn't really afford much, I eventually got an upgrade to a 386-SX running at a startling 16 MHz. These were the heady days of ISA slots and their 16 MB/sec of bandwidth. Sure, this was plenty for that time, as even the really high end soundcards couldn't take advantage of all that bandwidth. VESA Local Bus (VLB) was released, and it became the graphics and storage controller standard of the time. I had friends that bought AMD 486 40 MHz based CPU's and ran them on VLB motherboards, and everything seemed to run pretty smooth. Then came PCI. I was quite excited by the prospect of PCI, and I went to computer stores around town to price out systems that integrated this technology. One thing that I didn't expect was the absolutely poor implementation that PCI received on many non-Intel based chipsets, especially those still catering to the 486 crowd. PCI had a really rocky start, and the only really good chipsets supporting PCI were the Intel Triton and Neptune based products. Many of the Taiwanese chipset manufacturers had some real issues with PCI, and this lasted well into the Pentium era (who here remembers the "fabulous" Opti chipsets?). Finally things settled down again and most PCI implementations were very solid, though we were left with only a handful of chipset manufacturers (ALi, SiS, VIA, and Intel). Then came the next big change; AGP. This was a tough one for many of the chipset makers to implement correctly. Intel of course pushed the standard, and their products supported it correctly of course (since it was a standard designed by Intel). The first implementations from VIA and SiS had some real issues (VIA VP3 and SiS 5591 were really their respective first products supporting AGP), but oddly enough little ALi didn't show these problems. Later SiS seemed to fix their issues, but AGP was something VIA struggled with for quite some time. AGP 2X then morphed into AGP 4X, and again the Taiwanese chipset guys had more issues as well. AGP 8X was later released, and still VIA had problems (the first batch of Radeon 9700 Pro's wouldn't work on VIA's recently released AGP 8X chipsets). Finally, after many false starts, every chipset manufacturer had a product that would finally work with all AGP cards. AGP 8x was the defacto standard for several years, and it performed its duties more than adequately. Then Intel of course released PCI-Express. The first PCI-E products were the i9xx series of chipsets from Intel, and because Intel was the big push behind NGIO (Next Generation I/O- the working codename for PCI-Express), these products performed their duties perfectly fine. Now comes the shocking part. It seems that every chipset manufacturer that is supporting PCI-E is doing it correctly. So far I have seen no issues with compatibility, or of lower than expected performance. The rollout of PCI-E seems to be going with nary a hitch. ATI, NVIDIA, SiS, VIA, and ULi (formerly ALi) have all released their PCI-E compatible parts, and all of them seem to be working fine right out of the gate. From testing around the web, it appears as if there are no real winners and losers here in terms of overall performance. I think all of the chipset manufacturers and the PCI-E working group have done an excellent job with this technology rollout. This transition is much, much smoother than anything I have previously witnessed. With no readily apparent compatibility or performance issues, I think that the introduction of the PCI-E technology has been a huge success for the industry. With no real snags to hold it up, I would expect the shift to PCI-E will really accelerate from here on out. PCI-E is a great technology, and it will really help to release many of the bottlenecks that we were seeing on the old shared PCI bus. Networking and storage controllers are the first real benefactors of the technology, and once we start seeing more true native PCI-E graphics adapters we will start to see the software developers really take advantage of the bandwidth the PEG slot can give. So, now that memories of products like the VIA VP3 are long gone, we can move on to greater things. I was initially expecting to see issues with other 3rd party support of PCI-E, but the smoothness of the rollout has really surprised me. Now we only need to start seeing new components like soundcards being introduced in PCI-E form. With the increased importance of high quality audio these days, having more bandwidth and greater functionality will help to enable a more immersive sound experience. I am still hoping for NVIDIA to release some sort of SoundStorm for PCI-E!!! Though alas, my hopes will probably be dashed onto the cold, hard ground. Still, we can expect both Creative with their Audigy and VIA with their Envy technologies to eventually take advantage of PCI-E. We can all stop holding our breath now, PCI-E is here and is rock solid. Arpil 6, 2005 NVIDIA Opens SLI-Zone - Josh In a not totally unexpected move, NVIDIA has opened up its SLI-Zone, which is aimed at being both a gathering place of SLI owners, as well as a good source of info for those looking to upgrade to SLI. Some good info there for those looking, and there are a couple of really interesting prizes for those willing to do a SLI movie. The grand prize being a 2 x 6800 GT with MSI Neo 4 Plat/SLI setup! That ain't too shaby. NVIDIA took a lot of people by surprise with their nForce 4 Intel Edition products, as word on the street before this release pointed to the boards not performing up to par with the best of Intel. Well, while those reference boards should not be entirely indicative of what will be offered by NVIDIA's partners, I highly doubt that the boards that will be released will be any slower than the reference (it just isn't in NVIDIA's interest to have a super performing reference board, but then just leave scraps for its partners). Not only are the big three releasing products based on this (Asus, MSI, and Gigabyte), but the 2nd tier guys like DFI and Abit will have products out as fast as the big boys (unlike the original nForce 4 SLI for AMD release). So, we should see some innovation here, and Intel is probably ecstatic about the release of this very high profile, enthusiast level product. Not only do they now have a product that will compete with the AMD SLI platform in the extreme enthusiast market, but they also get to charge NVIDIA a licensing fee for every nForce 4 IE chipset sold. Since the release of the Athlon 64, the gaming world has been dominated by AMD machines. With the addition of the much hyped dual core products coming from Intel, this could give them a big boost in this small but lucrative market. I have two reviews that are right on the edge of being finished, the first being the MSI K8N NEO 4 Platinum (nForce 4 Ultra based), which has turned out to be quite a workhorse around here. In the past I haven't been a very big MSI fan, but that is starting to change. Their previous products often were well designed, but they took quite a few shortcuts in dealing with overall quality (especially in the power area). Well, times have changed and so has MSI. I feel they now offer some really top quality products that won't up and die without notice! I remember quite a few MSI boards that I used in the past that would burn out due to the above mentioned shortcuts, or just start to run unreliably. Well, looking at the components they are running now, it appears that they have changed their ways! The second product, and perhaps the more interesting, is the Albatron GeForce 6200 TC. This is the little train that could! While it is far from an enthusiast level product, it is a pretty impressive product for its price range. While there are a few caveats that I will be going over, it could be the perfect card for a wide variety of users. Something that I found quite nice about it is that it can run PS 2.0 titles very, very well at medium resolutions (with high quality settings enabled). This is not your average budget level card, and it is a stupendous jump for NVIDIA up from the GeForce 4 MX and GeForce FX 5200/5500 products. More later! April 5, 2005 Four Video Cards in SLI!!! - Josh This one is an exclusive! I am able to use four video cards in NVIDIA's SLI! Take a look here! NVIDIA nForce 4 for Intel - Josh Today NVIDIA officially unveiled their nForce 4 for Intel. The first iteration is coming out in SLI form, and the previews today all feature the SLI setup. The nForce 4 for Intel takes NVIDIA back to the two chip setup, which I believe is still the most functional and flexible of standards. The Northbridge handles all the PCI-E, FSB, and memory functionality, is comprised of 60+ million transistors, and is made on TSMC's 130 nm process. The Southbridge handles all of the usually southie functions, and is made on the 150 nm process. The two bridges are connected via HyperTransport (the irony, oh, the irony). The memory controller that NVIDIA has devised for this platform is very impressive. It features the true two channel controller and memory arbiter, as well as the DASP 3.0 functionality (data pre-fetch). In nearly every benchmark published so far, the nForce 4 for Intel beats out the top of the line i925X chipset from Intel. If you are an Intel enthusiast, this might be the product for you! It sports all of the features of the AMD based nForce 4 chipset, but also adds support for Azalia audio standard. This chipset should send a shockwave through the Intel ranks, as it appears to be an immensely competent product for the Pentium 4. At high resolutions its performance at SLI easily matches that of the AMD based product (though at lower resolutions when games are more CPU bound than video, the Intel platform does not perform as well as the AMD based one). Still, this looks to be a very solid product and will give NVIDIA a foothold in the larger Intel chipset market. While NVIDIA dominates the AMD sector, remember that AMD itself only holds less than 20% of the CPU market. Now that NVIDIA has the license to produce Intel based chipsets, their chipset business now has the added headroom to grow to huge proportions. If NVIDIA could capture just 25% of the Intel market, then their chip business would be roughly double what it is now. One thing I have to go back to is the memory controller. This controller not only utilizes the DASP functionality, but also introduces the QuickSync functionality. The Intel based chipsets have a single controller that utilizes the 128 bit path to the memory (hence "dual channel" 64 bit). But from my understanding it really only views it as one channel. The nForce 4 controller splits this up into two distinct channels, and the QuickSync is actually a channel arbiter. This allows for greater control and granularity of memory transactions. Latency may in fact be slightly affected by this arbiter, but it should be well covered by the increased overall efficiency of the memory setup. The controller utilizes DDR-2 memory, but it officially supports DDR-2 667. Apparently some users have been able to successfully run their high end DDR-2 DIMMS at DDR-2 800 on these boards. While NVIDIA may take a chunk out of Intel's chipset market, the very fact that NVIDIA is now making these chipsets should help to make the Pentium 4 a much more competitive product against the Athlon 64. It is a very good day for Intel enthusiasts, and we are told to expect retail nForce 4 for Intel boards by the end of April. My only question is if NVIDIA will go back to a two chip architecture for the Athlon 64, or does the lack of need for a memory controller on the chipset side negate any advantages a two chip architecture would bring? One would think that having a separate Northbridge for both AMD and Intel based products, but then be able to share out a handful of different Southbridges would give the motherboard manufacturers a greater leeway in designing products? I see what VIA has done with their multi-chip product line (many Northbridges able to work with several Southbridges) and that seems like the most efficient use of engineering and design time. Here are a couple of previews: Everything points to the nForce 4 for Intel being a success, but much of this success depends on how fast NVIDIA can get chips into the manufacturers' hands, and how easy it is to design a product around these chipsets. Now, the only thing I could really wish for was that NVIDIA had integrated more PCI-E lanes into their Southbridge!
If you have found this article interesting or a great help, please donate to this site.
Copyright 1999-2005 PenStar Systems, LLC. |
|||
