   |
|||
|
|||
|
:
News
: Archives
|
|
|
|
|
|
|||
|
|
|||
|
|
|||
|
January 28, 2006 Catching Up - Josh My good friend Joe James has moved on from Corsair and has joined up with Super Talent Memory. These guys have been in business for quite some time, but their name is not generally known in the enthusiast/high end market. Joe and the gang are hoping to change that. Super Talent just introduced two new dual channel memory SKU's: the X40PB2GC3 which is a PC4000 rated DDR-1 kit consisting of 2 x 1 GB DIMMS rated at CAS 3 at 500 MHz (250 MHz DDR), and the X32PB2GC2 which is a pair of 1 GB DIMMS running at PC3200 at CAS 2. Joe seems awfully excited to be there, and we should probably start seeing some reviews of their products next quarter. The stuff could be quite interesting to look into, and a good alternative to other memory manufacturers. They are also probably going to introduce new DDR-2 memories this next quarter in conjunction with AMD's new sockets. Jay Heo of BlueGears wanted to let me know that BG is no longer supporting the HDA series of sound cards. That means no more BlueGears branded/distributed X-Mystiques or the upcoming X-Plosion. Instead a new company called Auzentech is taking over distribution. As you can see from the site, Auzentech is not quite open for business. BlueGears is going to be concentrating on other products, but they haven't let slip what they will be doing yet. Hot Hardware has been pushing out the reviews lately, and they have the Squeezebox Wireless Media Player (which they tear apart quite nicely) and the PowerColor Theater 550 Pro PCI-Express TV Tuner. The latter is one of the first and only PCI-E TV Tuners around. Definitely worth a read. This past week was quite busy here with the release of the X1800 XT review, the Optimism editorial, and the Icemat review. There is also some brisk discussion of the Optimism article that you should all join in on. I am hoping that this next week brings just as much stuff! I am interviewing a potential writer for the site, and he lives in the Bay Area. This could be good, as he would be able to cover many of the tech conventions and releases in that area without me having to travel all the way from Wyoming to do it. More on that later. January 26, 2006 Icemat Siberia Headphone Review - Josh Some time ago (long time actually) Icemat Audio sent a sample of their Siberia headphones. At first I wasn't sure how to react to these, as they had a unique mix of quality, functionality, and price. So to get a better idea how these actually stack up in the real world, I enlisted the help of an audiophile friend of mine to review these. Ryan Rayda is an engineer and musician with unique qualifications. Ok, maybe not so unique, but he certainly has a good ear for sound. Here are some of his impressions:
You can read the entire review here. January 25, 2006 Exuberant Optimism - Josh Lately I have been reading a series of articles as well as analyst proclaiming that Intel will recapture the performance crown this summer with the release of the Conroe series of chips. I believe that in many ways this view is quite myopic, and in the end it will be proven false. I am not saying that the Conroe and its brethren will be poor performing, but rather these writers and analysts are not taking the entire picture into account. Conroe looks to be very good, and it is a product that Intel can be proud of. Unfortunately for Intel, it most likely will not wipe AMD off the map, nor gain back marketshare from its little green competitor. I do my best in this article to explain the reasons why. Here is a quote:
You can read the entire article here. January 24, 2006 ATI Releases the X1900 Series - Josh ATI has recovered the performance crown with their X1900 series of cards. Not only that, but there is good product available. While ATI has in the past delivered fast, new cards, they couldn't actually supply the top end. Now they are making a very positive change in that they have very good availability the first day these cards are released. This is about on par with the original 7800 GTX release, and of course the amount of cards on hand right now are thoroughly beating the 7800 GTX 512 release. I would imagine that quite a few GTX 512 pre-orders are going to be cancelled now that it seems the X1900 XT and XTX are slapping the GTX around a bit. Not only that, but you can find the X1900 XT at $549, and the X1900 XTX for $649. Compare that to the 7800 GTX 512 which retails around $749. I would say that the user would really have to like NVIDIA parts to pass up a cheaper, and faster board. Not only that, but you are paying for each one of the X's in the X1900 XTX! The R580 is the new chip the X1900 series are based on. This chip is fabbed by TSMC on their 90 nm line. It runs at 650 MHz in the XTX version, and its hardware isn't too terribly far off from the R520 that is in the X1800 series. ATI has done some work though, and there is one large change from the previous R520. Instead of having a 16 ROP, 16 Pixel Shader (2 ALU's a piece), 16 Z-compare unit, and 16 texture unit part ATI has expanded the pixel shading capabilities by incorporating 48 pixel shader units (still 2 ALU's). The R580 still has the 16 ROPS/Z-compare/texture units, so its overall fillrate is nearly the same as the R520 at the same speed. The pixel shader throughput though has tripled. ATI figures that as time goes on, most applications are going to rely on math vs. traditional texturing/fixed function pixel rendering. Shader ops will become far more common, and easier to implement with good results. The R580 is designed to address this rapid increase in math functions in pixel rendering. This increase in shader power gives the X1900 XT and XTX a good 10% to 30% increase in performance over the older X1800 XT, depending on the application. Games such as F.E.A.R. seem to get the biggest boost from the extra shader power. You can now find the X1800 XT for as low as $469, but that may be hard to justify when for an extra $50 the buyer can get a X1900 XT. Of course prices vary, but it doesn't appear as though the X1800 XT's will go down in price much, and we have basically seen the supplies of X1800 XL's drop to near nothing. ATI is fully phasing out the X1800 XT's, and the stop order for the R520 chip was made months ago. Something interesting is that the XTX version is only clocked 25 MHz more in the core and memory than the XT. Pretty much everyone is scratching their head on that one, as it makes little sense to pay another $100 for that small bit of performance. One thing that does come out though is that the XTX has slightly faster GDDR-3 and that it can potentially clock quite a bit higher (up to 1800 MHz theoretically). All we need now are the non-ATI tools to do such a thing. ATI is releasing three X1900 cards today; the XTX, XT, and the CrossFire Edition. The CF edition is clocked the same as the XT (625 MHz core, 750 MHz GDDR-3). So when a user plugs an XTX in with the CF edition, it downclocks the XTX. So, if you are looking to invest in CrossFire, it makes more sense to get a XT edition. ATI has also released the All-In-Wonder version, but this card won't be available in quantity for some weeks. ATI has done a great job releasing this card to widespread availability. It also seems that in the majority of applications the X1900 XT and XTX are the standalone performance winners. ATI has taken back the crown, but NVIDIA obviously won't be standing around twiddling their thumbs. Within the next month we should hear more about the G71 based cards that look to be turning up in March. I am curious if NVIDIA will again hold back until they can hard launch this product, or if they want to take some of the shine off of the X1900 series by doing a regular release to media and then providing availability up to a month afterwards. You can imagine that this discussion is going on right now at NVIDIA. The G71 does appear to be more traditional in its pipeline configuration than ATI's latest, but it is looking like the G71 will feature a full 32 pixel shader pipes with 2 full ALU's a piece. It also looks to be clocked around 650 MHz to 700 MHz, which will give the ATI cards a severe run for their money. Time will tell, but for now ATI is savoring the engineering victory that they have achieved this day. Unfortunately, the review deities did not smile down upon me, and I received no cards for review. Instead I will point you towards other reviews popping up on the web. January 20, 2006 Sapphire Radeon X1800 XT Review - Josh Sapphire, being the close partner to ATI as they are, usually gets the first 3rd party Radeon chips. This is not surprising as Sapphire's parent company actually makes the majority of ATI branded boards on the market. Sapphire has put together a very solid bundle and card for the X1800 XT offering, and they have given me the chance to review it here. Should the idea of buying this card in the face of the upcoming X1900 scare you? I personally don't think so as the X1900's won't provide a significant amount of performance over the X1800 XT, and the X1800 XT will be quite a bit cheaper until product is all sold out. Here is a quick quote:
To find out more about this great little card, read here. January 19, 2006 AMD Posts Record Revenue - Josh After Intel's disappointing earnings two days ago, many investors were waiting to see if the same bug affected AMD. It did not. AMD posted $1.84 billion in revenue, and after a one time $110 million dollar charge due to AMD spinning off Spansion, AMD posted a $94 million profit for the 4th quarter. This is up significantly from Q3 where AMD had revenues of $1.56 billion with a profit of $76 million. Without the one time $110 million charge AMD would have had a record profit of $204 million. Now that Spansion is totally spun off into its own entity, their earnings will no longer be included with AMD's. This means that this current quarter will appear down overall from the previous quarter's result, but overall the company will be a bit more healthy. The processor group alone accounted for $1.35 billion this quarter, and AMD is looking at Q1 this year (typically a soft quarter for this industry where "seasonal changes" usually mean that revenues and profits are down) as either flat or slightly down. They are also being conservative with that estimate. AMD has several things going for it. AMD is selling nearly every good die coming off of its production lines at Fab 30, and we are already seeing near mature yields off of Fab 36 product lines. AMD expects first shipments of Fab 36 parts in early February. AMD's channel inventory has been flowing very smoothly, and AMD reports no real excess inventory to speak of. This says a lot of things. Not only is AMD selling a lot of parts, but they are handling product placement and pricing near perfectly for the market. Instead of merely saying, "This is what the market needs" they are obviously looking around at what is selling, what price points are being hit, and how their products are compared to the competition. Parts that are no longer selling are phased out, and the channel is not left with a lot of excess inventory due to good sell-through. Intel is not in that position. Apparently Intel has a lot of excess inventory that is not being sold, and their production lines are underutilized. I am guessing that this situation is due to Intel really pushing 65 nm, increasing its overall production capacity, yet not having any real growth to utilize that extended capacity. This may very well change with the introduction of the Conroe family, but I do not believe that these new cores will be the panacea that many analysts are hoping for. With Fab 36 coming online we may see more companies utilize AMD processors due to merely the idea that AMD now has a lot more capacity to meet customers' needs. Dell has always claimed that the reason they have never used AMD products is that AMD cannot promise to fulfill the inventory needs of Dell. Intel can make that promise, and in fact handle all the inventory work of Dell (Intel sells directly to Dell, and Intel makes sure that Dell gets everything they need). Rumors have been swirling around about Dell offering an AMD product, and while I cannot vouch for this rumor, several things do stick out that might in fact point to Dell adopting some AMD products. The first is that AMD will in fact have more production capacity very soon, and I wouldn't doubt that Fab 36 parts are already in the market. Next we have seen Dell break away from the all Intel product offering by going with the NVIDIA nForce 4 for Intel chipsets. Previously Dell has only utilized Intel chipsets to go with the Pentium processors, and even with Intel pushing the ATI chipsets for the budget integrated sector, Dell is still using the Intel integrated chipsets. So, with that small break we can see that Dell might be more willing to utilize AMD parts because they already have a good relationship with NVIDIA on the chipset side. NVIDIA provides a wide range of chipsets for AMD from the Professional versions (server and workstation) to the bottom end (6100/410 series). So, while it is not a given that Dell will actually go with some AMD parts, the stage is set for AMD to break into this final bastion of Intel Only. With the increased production we will also see other companies such as HP offer more products based on AMD parts because AMD has the ability to address larger product lines from the major manufacturers. Performance and power consumption are two hallmarks of the Athlon 64 lineup. The desktop and server processors are all hailed as the best in their class, and the Turion is starting to make a big splash in the marketplace. This is the next market that AMD is focusing on and they have a series of programs to make the Turion more competitive with the Centrino family of products. This is an area that Intel is much more competitive in due to the Pentium M series of products (which are quite good). With the introduction of the dual core Yonah, Intel is set up to still be the dominant player in this area. There are some issues still with the Intel products though. The biggest is the lack of a 64 bit part in the notebook sector. With Vista coming out this year, and with notebook product cycles so much longer than desktops, not having a 64 bit notebook part could hurt Intel in the high end of the market for the first half of this year. This will be a place where AMD will start to gain marketshare, as those looking to buy a high end machine might just want the ability to upgrade to a 64 bit operating system. Also, the older FSB architecture for Intel is slower, and eats more power than AMD's HyperTransport. Not only that, but the one-two punch of the integrated memory controller and HyperTransport for the Athlon 64 architecture makes for a very compelling and efficient architecture, and one that is easier to implement in the cramped confines of a notebook than Intel's older architecture. Add to that two very good integrated chipsets from both ATI and NVIDIA, and we are starting to see a swing in the notebook sector towards AMD. Process technology is a two edged sword at the moment for AMD. Their 90 nm process is simply world class, and their ability to tweak it and get outstanding results is second to none. The latest FX-60 processor, though it runs 200 MHz faster than the older X2-4800, actually pulls slightly less power and runs slightly cooler than the initial 4800 products did 9 months ago. For a process that many had relegated to the "too old and too big" dustbin, AMD has been able to tailor it to their needs. We look at the power and thermal characteristics of the single core Venice and the dual core Manchester, and the results are simply astounding in terms of what AMD has been able to do. I would expect to see the first dual core Turion products released early this Spring using the DDR-2 dual core/512K per core design. From all indications, these cores will have better thermal and power characterstics than the current DDR based Athlon 64's, as well as having a better speed bin mix. Add to that the extra power savings of using current generation DDR-2, the Turion series of chips could receive a lot of attention during this next notebook product cycle. The switch to 65 nm is well behind where Intel is at (they are actually shipping product), but AMD and its process partners simply do not have the money to throw at that problem that Intel does. AMD is expecting their first 65 nm parts to come off the line in Q3, but that is not a given. AMD has claimed that they have hit every milestone, and their results are good for this process change. AMD has partnered with IBM and Infineon to develop their 65 nm process, and the extra capital that is injected in this process will help them to better compete with Intel in the future. AMD's process engineers are some of the best in the business, and with the ability to share the load with the top IBM and Infineon guys, this should make this process switch a lot more cost effective than AMD going at it alone. Currently the 65 nm process is not ready for production, but I think AMD has a very strong chance of meeting its expectations of a Q3 release. With AMD being so conservative about their business as of late, I would actually expect full production of 65 nm parts to start mid-Summer and be released shortly after Intel introduces their Conroe series of products. For the next six months AMD will continue to gain marketshare and mindshare through their outstanding products, and the switch to DDR-2 with these new cores will help in terms of performance and power consumption as well. While AMD does not have the 65 nm advantage at this time, Intel currently does not have their Conroe based parts to really take advantage of the situation. Once Intel does release these parts, it appears as though AMD will have an answer for it shortly after introduction (one to two months). Needless to say AMD will have product ready for the important Holiday season product cycle in late 2006. Post your thoughts here. January 18, 2006 NVIDIA Releases New Products and Futuremark - Josh The past two days have been chock full of NVIDIA product releases. The first is the release of two new Intel supporting chipsets; the nForce 4 SLI XE and the nForce 4 Ultra for Intel. The original SLI for Intel featured 2 x 8X PCI-E slots, and did not support the latest 8x0 and 9x0 series of dual core Pentium 4 chips. The high end SLI X16 nForce 4 changed that issue, as well as provided up to 32 PCI-E lanes for graphics. The new SLI XE now supports the new processors from Intel, and it has most of the nForce 4 features as well as the ability to support 2 x 8X PCI-E lanes for graphics. The Ultra of course can't split the lanes for SLI, but otherwise it has the same features as other nForce 4 chipset. The big difference between the SLI XE/Ultra vs. previous nForce 4 products is that they do not support the Active Armor hardware acceleration. Two big selling points for these chipsets is the inclusion of the HD Audio as well as the support of DDR-2 667 (and beyond... unofficially). These will help NVIDIA get a bit farther into the Intel market since Intel has shifted its chipset focus more to the high end. There is still not an integrated graphics version of the nForce for Intel chips yet, but this could be a growth area for NVIDIA if they decide to pursue it. Due to the reliance of the GeForce 61x0 chips on the integrated memory controller on the AMD Athlon 64 products, it would be non-trivial to port that design over to the Intel side which would then use the NVIDIA designed DDR-2 memory controller. The other announcement is that of the GeForce 7300 GS series of cards which will start to replace the 6200 TC series of chips. The 6200 TC was a very solid budget part for NVIDIA, and the ability to utilize main memory in an efficient and fast manner led to some very inexpensive cards. The 7300 is the first desktop 90 nm chip that NVIDIA will produce, and NVIDIA characterizes their switch to 90 nm as "very good". The 7300 runs at 550 MHz core, features 2 ROPS, 4 pixel shader pipelines (2 ALU's each) based on the G7x tech, 3 vertex shaders, FP-16 filtering, and finally support for the FP-16 buffer (for HDR effects). The previous 6200 had the FP-16 buffer disabled due to space/performance reasons. These cards will feature a 64 bit memory path to main memory, and will come with 128 MB or 256 MB of onboard memory (DDR-2 clocked at 350 MHz). The ability to support 256 MB onboard will allow it to use TurboCache to make it appear like the card has 512 MB of memory total. Another area of interest is PureVideo. We haven't received many details of the upgrades that PureVideo will go through in the coming months, but unlike the 6200 TC the 7300 series will support PureVideo acceleration up to 1080i (720p used to be the maximum for the 6200 series) and includes many of the disabled PureVideo features in the 6200 (high quality de-interlacing, etc.). This will be a good upgrade to the 6200 in terms of features and performance, and it will probably be a bit faster than the competing X1300 series of cards from ATI in 3D applications. One big question is how well it will accelerate H.264 video and at what maximum resolution it will do it to. Overall though, this is a significant upgrade of the budget space, but NVIDIA is not exactly "hard launching" this product. Product will not be available in the States until early February. This should not cause a major uproar because the vocal enthusiast crowd do not necessarily care about the budget products. Still, this is a good minimum performance product for game developers to aim for. In other news Futuremark has now released their 3D Mark 2006 benchmark. This has three upgraded tests from 2005, as well as one new test. Futuremark has changed around the testing procedure a bit, so now it reflects overall system performance as well as pure graphics. It looks like a pretty significant upgrade in many ways from 3D Mark 2005, and it will definitely stress out even the most high end cards at this time. January 13, 2006 BlueGears X-Mystique 7.1 Soundcard Review - Josh For the first review of 2006 I have posted my impressions of the BlueGears X-Mystique 7.1 sound card. This is one of the first products since the NVIDIA SoundStorm to support hardware Dolby Digital 5.1 encoding. The card also has a host of features to add to the DD 5.1 support, and overall it is an impressive package from C-Media and its partners. Feel like trying a sound card away from the Creative brand? Then start looking here.
You can read the entire review here. January 11, 2006 AMD FX-60, Intel and Apple, and other tidbits - Josh CES is now past us, but the news keeps churning... AMD released the FX-60 this week, and while it isn't anything terribly surprising, it does have a few interesting wrinkles. The FX-60 is a dual core Toledo chip running at 2.6 GHz. It is still based on 90 nm process tech, but these chips have been shown to actually run cooler and pull less power than its older X2 4800 sibling. This means that AMD has really been working hard on their process to keep their chips running cool even at higher clock speeds. While the FX-60 is in fact 200 MHz slower than the FX-57, the addition of another core makes it almost as fast in most applications that are single threaded, and it simply blows the FX-57 away when multi-threaded apps are used. The FX-60 appears to clock about as high as 2.8 to 2.9 GHz without much issue. It also looks as though the FX-60 will be the top of the line chip for AMD for quite some while, though it will receive a refresh this spring when the AM-2 sockets are released. AMD might be able to squeeze 2.8 GHz out of their current 90 nm process with a dual core product, but I wouldn't count on it. It appears as though the next significant jump will be when AMD releases their 65 nm process based parts. For the time being, the FX-60 is going to be tops for quite some time, and it will be interesting to see what sort of changes to performance come about when the AM-2 socket is introduced. IBM and Infineon are working very close to AMD to achieve good results with their shared process tech. AMD just doesn't have the funds to go at this alone, so they have chosen to create partnerships with other companies around the world to share process tech, but of course most of these companies are not direct competitors with AMD. Intel continues to go at it alone when it comes to process tech, but when they have billions in the bank it is not surprising. Though sometimes I believe that Intel sometimes suffers badly from NIH (not invented here) and it has a negative effect on some of their products and technology (eg. re-inventing the wheel, dismissing tech advances out of hand, etc.). Many of the reviews out there about the FX-60 compare it directly with the Intel 955 Extreme Edition. Most of these reviews are not terribly friendly to the Intel part, but that isn't a shock. While the performance of the 955 is quite good, the FX-60 usually one ups it. The real kicker is the power and heat numbers between the two products. The FX-60 keeps its thermals under control at all times, and it really doesn't use much more power than its predecessors. The 955 on the other hand is re-writing the book on heat production and power consumption. This is a big stickler, and it is an issue that many users and OEM's are quite unhappy about, especially Dell. As a current Intel-only manufacturer, Dell has to go to great lengths to introduce these high end parts and yet achieve good thermals as well as low noise. The 955 essentially requires a motherboard with a 5 phase power supply just to keep the power flowing smoothly and not causing the MOSFETS to glow red due to heat. Remember that the 955 is a 65 nm part. Now contrast this to the very quick and cool Yonah processor. It really is amazing that Intel stayed with the Netburst architecture this long. Something that really sticks out when comparing Intel and AMD products is the overall memory efficiency between the two. Intel still relies on the old GTL+ based FSB. The original GTL+ was released with the Pentium Pro, and then made it to the Pentium 2 and beyond. When Intel went to the Pentium 4 they changed it a bit so it could handle 4 bits of data per cycle, basically making it a quad data rate signal. But unfortunately some of the same limitations from the Pentium Pro days are still present. The biggest issue is that Intel cannot do synchronous memory read/writes. The command to do a read is sent, and everything waits until that command is finished, and the same goes for write commands. AMD on the other hand utilizes a full duplex system in both the onboard memory controller as well as the HyperTransport links. Something else to consider is that on the Athlon 64 the memory controller is running at CPU core speed, so many transactions are significantly faster than the Intel memory controllers which are located on their chipsets that run at 266 MHz on the high end products, and 200 MHz on most of their other products. Yonah on the other hand has the memory controller running at 166 MHz. These are huge advantages to AMD, and I think they can really help the Athlon 64 compete very well with the Conroe series of chips. Conroe, Woodcreset, and Merom will all still utilize the aging GTL+ FSB running at 200 MHz (266 MHz for the high end). Intel will eventually be going to more of a HyperTransport like system, dubbed CSI. Intel is already bragging that their implementation will be better and faster than AMD's HyperTransport... but take that with a grain of salt. The first CSI products will not appear until at least 2007, and it could be 2008 by the time it hits the desktop and notebook market. AMD does not stand still on HyperTransport, and the third version of HT should be implemented next year. So who knows what changes will happen in the next two years with AMD processors and their HyperTransport links. So overall I think it is a bit silly for Intel to take such a stance at this point, but then again we are talking about marketing. Apple and Intel are showing off the first fruits of their partnership, and these new iMacs and MacBook Pros are certainly impressive products. They feature the Intel Yonah series of chips, both Core Duo and Core Solo. It now appears that Apple will skip on any Netburst chip, and go straight to the better performing and lower power Pentium M series. Philosophically the deal with Intel doesn't make much sense to Apple, especially in light of the AMD Athlon 64 series. Once we consider many real world situations, the partnership with Intel makes a lot of sense. Intel can provide an entire ecosystem of parts for Apple (processor, chipsets, wireless, etc.), as well as be able to insure that Apple receives the supply of chips they need. AMD is not in that position, especially considering they sell every Athlon 64 core that comes out of Fab 30. With Fab 36 coming on line late last year we are now probably starting to see the first trickle of 90 nm products from that fab, but it won't be until 6 more months or so that we will see a larger utilization of that plant. By that time we should start to see the first trickle of 65 nm parts from AMD as well. By the end of 2006 AMD shouldn't be as production constrained, but since Apple is starting to release their products now, later this year is not good enough. We also have to consider the marketing money that Intel can throw at Apple. In a business sense, partnering with Intel is a great deal for Apple, but they have had to give up a lot in terms of image and technology as well. Remember, the Yonah series of processors is only 32 bit, which is a step back for Apple since their latest series of G5 products were all 64 bit. As you may or may not notice, Apple isn't talking about 64 bit anymore. So, from a technology standpoint Apple would have been better off with AMD, but from a business and marketing angle it makes far more sense to go with Intel. Still, this partnership will improve Apple's stance in the industry, and they will be in a better position to supply better parts at better prices. When the next gen of Intel parts hits, then Apple will be able to refocus on the 64 bit angle. ATI was pretty quiet at CES this year, but I think they wanted to save their thunder for later this month. Rumor has it that ATI will be releasing their R580 on the 24th of this month, and I see nothing to contradict this. The R580 will be a pretty significant jump from the R520, both in terms of performance and power consumption. ATI most likely will take back the performance crown with this product, as it really doesn't appear as though NVIDIA will be able to push out the G71 until late February/early March. Initial samples of G71 boards are out and about, but it doesn't appear as though NVIDIA can provide enough product for a hard launch until much later. ATI on the other hand is said to be shipping a lot of R580 products to its partners, and their launch later this month looks to be a true hard launch for ATI. This really is a first for the company, as most of their previous launches had little or no product on the shelves until a good 2 months had passed. This will be a welcome change for ATI and its partners. NVIDIA will of course not stand still, but from all appearances they will not be able to hard launch the G71 until later. To offset this, I believe NVIDIA will refresh their 6600 and 6200 series of cards with the 7600, 7400, and 7300 series of products. While these will not overwhelm the release of the R580, it will provide a welcome refresh for the budget and midrange markets. The 7600 will be faster than the 6600 series, and these cards will most likely overshadow the X1300 and X1600 parts from ATI. A little tit-for-tat here, but every bit of marketing counts to these guys. January 6, 2005 NVIDIA has certainly been busy at CES this year. The big news is the Dell/NVIDIA designed Quad SLI package that is going to soon be available only from Dell with their XPS 600 series of machines. It is unknown if this tech will reach DIY hands, but it would be nifty if it could. It requires the use of a motherboard with a full 32 PCI-E lanes for graphics, which is currently supported by the nForce 4 SLI 16X chipset. The graphics cards themselves are very unique. Each "card" is comprised of two separate PCB's which house a 7800 GTX 512. Two of these "cards" are then used together, and two separate SLI bridges are installed. This custom system provides a HUGE amount of bandwidth and would enable a solid gaming experience with the new Dell 30" LCD. 2 GB of video memory doesn't hurt either, and NVIDIA has enabled a new AA mode- 32X AA.
This is a conceptual slide of the system, but apparently they are showing the real thing at CES at the moment (I believe PC Perspective has an actual picture of the system). Because of the length of these cards, NVIDIA was able to spread out the cooling as well as the power circuitry. No performance numbers are given so far with this setup, and we don't know how well the latest drivers are at providing good efficiency with 4 GPU's. One things is for certain... this will be an expensive option for Dell to provide. If you figure that on the retail market a 7800 GTX 512 is around $749 a piece... four of them cost almost $3000. Also, since NVIDIA will be releasing the G71 later this quarter, I would expect Dell to offer those chips instead of the older G70 that is currently available. It does look to me though that these will be externally powered units, and I believe the plug-in is that little slot under the cooling grills on the backplate. Needless to say, this is at the least an interesting engineering exercise, and it should still be a very fast solution. NVIDIA has also announced three new notebook GPU's. Previously they released the GeForce Go 7800 GTX, but now they have added the GeForce Go 7800, 7600, and 7400/7300. The Go 7800 appears to still be based on the G70, but with what appears to be a lower clockspeed and 20 pixel shader pipelines. The 7600 looks to be a new GPU, namely the G73. This is the update to the very successful 6600 lines of chips (NV43) but with G70 pixel shader enhancements as well as more vertex shaders. It still appears to be an 8 pixel shader design, but it will have enhanced performance over the previous generation of chips. The 7400/7300 is based on the G72, which will replace the 6200 series of chips. It appears to be a single quad architecture with the G70 enhancements. Both the G72 and G73 are based on TSMC's 90 nm process. Clockspeeds and power are not yet released with these products, but it seems that NVIDIA has yet again introduced their next generation products for the desktop in the notebook sector first. This likely means we will start to see the midrange and budget replacements for the very successful 6200 and 6600 series early this quarter. The replacement for the 6600 GT/DDR-2 series of cards should offer more performance and more capabilities, but still be priced under $199 at introduction. NVIDIA is hoping that this will be a true successor to the 6600 series, as the NV43 is still selling strong after 1+ years of being on the market. Two thing that will really help these cards along is that it appears as though NVIDIA has done some work on the memory controller to increase efficiency, and there are now very good supplies of high speed GDDR-2 and GDDR-3 modules. We can expect bandwidth for the top products in the 7600 series to hit 16 to 18 GB/sec with only a 128 bit memory interface. NVIDIA is really putting a lot of energy into their notebook lineup to help sneak marketshare away from ATI and Intel. This is a smart thing, as the notebook market is growing far more rapidly than the desktop market is. With the upcoming G71 as well providing a new top-end notebook part in the near future, NVIDIA will have a full spectrum of products supporting laptops from the GeForce 6100 series to the "yet to be announced" G71 notebook variant. Since notebook product cycles are a lot longer than desktop, we can expect to see many notebooks supporting NVIDIA graphics technology to be released by the end of February. At that time it will be interesting to see the numbers for notebook graphics marketshare, and if NVIDIA's aggressive product schedule is in fact working. Something else I have heard mention of is NVIDIA will upgrade their PureVideo capabilities. Little is known about this, other than they will support H.264 acceleration in hardware by using the programmable pixel shading hardware. Other quality enhancements appear to be in the works as well. January 5, 2006 The New Year - Josh Let me state this now... this is going to be a crazy year in the computer industry. Graphics, processors, and sound will have much more competition between the players than perhaps any other year I can think of. On the CPU side AMD and Intel are just wailing on each other. While AMD rules the roost in the desktop, they have solid notebook products based on the Turion processors, as well as the Opteron. Intel still has 80%+ of the market in overall sales, but it is starting to be chipped away by AMD and there are of course rumors flying that Dell will finally release an Opteron based server this quarter. This year we will see Intel totally revise their processor architecture and get rid of Netburst altogether. This should give Intel a more powerful and sustainable core in the 2nd half of this year, and AMD will have their work cut out for them to keep gaining on the processor giant. This doesn't mean that AMD will stand still, as their next gen of 90 nm processors will be released this Spring with DDR-2 support as well as some other internal tweaks to give a boost in performance as well as stretch out the clockspeed a bit. The shift to 65 nm looks to be going well, and AMD says they are hitting their milestones as expected. Intel has already released their first 65 nm products, and consumers should be able to start grabbing these shortly. AMD should continue to hold the overall lead until Intel releases the Conroe variants, and then it is going to be a major fistfight after that. AMD will still have the platform advantages over Intel (HT, integrated memory), as the Conroe and its stable of friends will continue to use the same old FSB that Intel has had for ages (which honestly has some major disadvantages and really needs to be chucked). AMD's AM2 based processors will be the hot item this Spring, so those looking to upgrade soon should probably hold back a bit until these products are released (wait 3 months). On a side note, apparently more than one reviewer has had the runaway thermal issue with the 955 Extreme Edition. These things run really hot, and it appears that if the VRM on the motherboard isn't as solid as it is supposed to, these chips go into shutdown mode (100C+). While it is true that the 955 EE is a brand new product, and production should improve on them, it is just a design that is incredibly power and heat unfriendly. AMD on the other hand has the X2 4800+ that runs around 48C at load, and is more than a match for the 955 EE. I know which product I would buy! On the graphics side ATI and NVIDIA will start this first quarter with a bang. ATI looks to be first to strike with the R580, which appears to be in the form of the X1900 series of cards. This will be clocked higher than the current R520 chips, and it should have a solid performance increase over current X1800 cards. Looking at the overall design (or what I know of it) I don't know if it will be the 2nd coming that many expect of it. Don't get me wrong, it will be a great chip that will perform quite well, and beat out the X1800 XT and GeForce 7800 GTX 512, but I don't think it will be much more than 10% faster in most current games and other applications. Yes, it does appear to have 16 ROPS and 48 pixel shaders, but I honestly wonder how much work each shader does as compared to more traditional designs like the X800 series and the 7800's? Many are just saying that R580 is almost like four RV530 chips put together. It is hard to say at this time how the entire package will work, but ATI is moving towards more decoupled designs while NVIDIA continues to embrace more traditional pipelines. For example, each shader pipeline for NVIDIA features two full ALU's, one of which handles texturing duties. So, at any one time these ALU's can do two shader ops or a shader op and a tex op. ATI's X1800 based on the R520 chip also appears to be made of 2 full ALU's and a tex unit, so it can sustain a lot more shader ops while still doing a tex op. The RV530 and R580 don't appear to be able to do that. ATI decoupled the tex units from the shader pipes altogether, and it appears that each of these shader pipes now only contain one ALU. This does mean less work per shader pipe, but with ATI's focus on threads, it can work out to be a more efficient design with more instructions being worked on at one time. On the other hand we look at what is to become the G71 from NVIDIA, and it appears to be a 32 shader pipe monster that is very traditional. That means that NVIDIA most likely have yet again worked to optimize per shader performance and will continue to have two full ALU's per shader pipe, for a total of 64 ALU's and 32 tex units. It also appears that NVIDIA has done some ROP work as well, but these will still most likely be limited to 16. Still, in both shader heavy apps and older texture based games, the G71 running upwards of 700 MHz will still be a beast and a tough product to contend with. It appears that NVIDIA is not nearly as close to releasing their product as ATI is at this time, but many are saying that NVIDIA will push up their timetable to compete with ATI. We will hear more by the end of January. With Windows Vista to be released later this year, both ATI and NVIDIA will again be showing off brand new architectures towards the end of Summer 2006. The G80 and R600 will be big steps up from current chips. Who said that graphics development was slowing down? As for sound... Creative will continue along with their new X-Fi series, but C-Media looks to have a couple of chips on the horizon that will at least give the giant a run for its money. Blue Gears (in the US) will be releasing some new cards based off of new C-Media chips, and will support better DSP functionality as well as Dolby Digital DTS encoding. I have been testing the older X-Mystique card for some time, and I have been impressed with the design and performance. This next card looks to build on that. C-Media also recently signed a deal with Intel, so that will be more cash in the company's pocket, which will lead to more products. This year could be one of great growth for C-Media, and we all know that Creative needs the competition. Happy new year!
If you have found this article interesting or a great help, please donate to this site.
Copyright 1999-2006 PenStar Systems, LLC. |
|||
