   |
|||
|
|||
| :
News : Reviews : Editorials : About : Contact : Advertising : Privacy : Links : Forums
|
|
|
|
|
|
|||
|
|
|||
|
|
|||
|
March 19, 2005 X800 and X800 XL Pricing Confusion - Josh Back on March 7 I talked about the MSRP's of the X800 and X800 XL products from ATI, and how much you can actually buy them for. The discrepancy is actually quite large, and I am not the only one who has commented on this. Scott over at the Tech Report has done a pretty good writeup on the current situation. I really am not sure what ATI is thinking here, or why they made the decisions they have. Did they feel that their products were not competitive at each price range with NVIDIA, and decided to put artificially low MSRP's on their products? Is this PR gone bad? I know that I get emails from NVIDIA PR quite often pointing out every flaw that ATI could possibly have, and this type of behavior is quite common and the amount of jockeying between these two companies is always intense, but this is one time where I am in agreement with NVIDIA's PR. I still think that ATI's lineup is still really shaky in terms of market focus. They have a lot of products out there that overlap, and when I do my X800 Pro vs X800 XL comparo, we will see a perfect example of this. The X800 Pro's typically cost a lot more than the X800 XL's, but the X800 XL usually beats up on the Pro. Then we have the situation where the average price of a X800 XL is around $60 higher than the MSRP on the product (though there are exceptions to this, such as the Gigbyte X800 XL which can be found online for around $290). We also have the X800 sitting well above its $199 mark, but part of this is related to most of those products coming with 256 MB of memory, which drives the price up a couple of notches. Still, there are very few of these products that even hit $249 (most average around $269 and up). Then we have the Asus versions of all of these cards, and I must admit that Asus must be very proud of their products as they often have an EXTREME premium on them (the Asus X800 XL version is around $430... nothing like having to pay an extra $131 over MSRP). ATI is also suffering at the hands of reviewers and competitors with the X700 series. When the X700 first was announced, the XT was going to be the high end part. It never showed up. ATI decided that for once there was too much overlap in their product schemes, and cancelled the XT version. This was probably quite smart on their part, because the X700 XT would really be performing about the same as the X800, and they would both exist at the same price point. Now, originally spec'ing out this part, and giving it to reviewers was probably a bad bit of planning. Now, having said all that, I have really enjoyed using many of ATI's products. I have a X800 Pro that happily sits in a test machine, and I have had that one for about 10 months. The X800 XL that I finished testing on is a very solid and well performing part, especially for its price. Overall it matches the GeForce 6800 GT in nearly every area, and it retails for about the same as the 6800 GT parts. Now, I am not saying that this is a huge conspiracy that ATI is trying to rip off gamers and enthusiasts, as their products are very good and match everything from NVIDIA pretty much step for step. My issue is the amount of confusion that ATI has sown into the marketplace with these dreamy MSRP's for their products. The X800, X800 XL, and X850 series were announced and released in late December, and only now are we really starting to see great supplies of these cards, but most often their prices are nowhere near MSRP. My main vid card distributor lists the X850 XT PE for $605 (that is my wholesale price), while that card is supposed to retail at $549. This same kind of markup is true up and down their line (except for their X700 series, which hits their pricepoints dead on). Now, one piece of good news that just came my way is a new X800 Pro PCI-E SKU that looks to be very interesting. This is a true X800 Pro PCI-E card, with VIVO capabilities, and is available at wholesale for $310. This means that these products will soon be available on the retail market for around $325. Now, while not as fast overall as the X800 XL, this is a VIVO version, which might very easily be converted to a X850 XT, and can probably be clocked up to X850 XT PE speeds. The possibilities for this product are quite exciting. Definitely something to keep your eye on for a future upgrade. March 17, 2005 New of Interest - Josh The kind people over at Tech Report have taken a good, long look at the Abit A8N Fatal1ty motherboard. This little number sports many gamer oriented features not normally found on a motherboard. The RAM is actively cooled by a "riser fan", the mosfets have their own heatsinks as well as a ducted fan that runs cool air from outside the case over them, red LED's stationed around the board, and not to mention all of the other regular tweaks Abit has going for them. This nForce 4 Ultra based board looks to please that niche of computing enthusiasts! ATI has released the X700 Mobility, and HardOCP is one of the first groups to get their grubby hands on the technology. Based on the 110 nm process at TSMC, this X700 product is aimed at the thin and light sector of notebooks (as compared to desktop replacements). It still has some awesome pixel pushing power though, and Brent J. at the [H] hammers it for all it is worth. Anand and the boys have several articles that are well worth the time to read. The first is coverage of the PhysX PPU from Ageia. The next is their take on multi-processing for the desktop and gaming. The final is a solid look at the Cell Microprocessor. They certainly are busy over there at Anandtech. My pseudo friend Ryan over at PC Perspective has taken a look at AMD's dual core plans for the rest of this year. While Intel was all dual core at their IDF several weeks ago, AMD has quietly been progressing on their own roadmap and we expect to see dual core Opteron chips before Summer. AMD is expecting their first desktop dual core product to be out sometime in the 2nd half of this year. Since Ryan brought up dual core, I thought I would say a couple of things here. While Intel has the R&D resources to really push their dual core strategy, and to get products out in an accelerated manner, AMD has really taken their time with their design. One aspect that many people seem to overlook is that Intel has to release a new chipset that supports their dual core chips, while AMD's dual cores can be dropped into current Socket 940 and 939 motherboards with a BIOS update. This is all because of the flexibility of the HyperTransport bus that AMD uses. The integration of the AMD dual cores into the processor package also appears to be a lot more well engineered than Intel's internal linkages, and this could pay off in greater efficiency in data handling between the two cores. Still, it is nice to see both companies working hard to provide the average user with more power without having to install a phase change cooling system into every computer sold. It is good that we are getting away from the clockspeed race, and working more on IPC and total system efficiency. Thats all! Time to go to my real job! March 16, 2005 PenStarSys.com Hardware Workshop and LanFEST 2.1 - Josh On February 25th and 26th, the UW PC Gamers hosted their LanFEST 2.1 event. This major gathering of Wyoming gamers turned out to be a very good success, with plenty of sponsorship and attendence. This is an event that is planned to be held at least twice a year with major events, as well as smaller events throughout the year. During LanFEST 2.1 I was given the opportunity to give a workshop on computer hardware. While there is plenty of room to grow for this event, it was still a success.
You can read all about it here. March 15, 2005 A Sad Day for Gaming News - Josh I just found out moments ago that Billy "Wicked" Wilson of VoodooExtreme and GamingGroove fame has passed away. I remember setting VoodooExtreme as my homepage many years ago when Billy was in full force there. He was certainly a driving factor in the gaming business, and his reporting and writing had a sense of style that no one has been able to imitate since. Leaf blowers and Vaseline were never the same after Billy got a hold of them! Billy was often very helpful to beginning writers and webmasters, and I know that when he was at VE he always did his best to post links to my then new site. He was also very available to people to talk to, and I bet he spent hours replying to emails from fans and other writers/editors. Billy passed away last night at 9:50 MST in Salt Lake City (yes, it was always a shocker for me to see that he stayed in SLC, his character just seemed out of place in that city). I will always remember Billy for his outrageous sense of humor, his dedication to gaming, and his ability to communicate with others and make them feel right at home (as well as important). He was definitely an innovator in the web reporting business, and his legacy will be felt for a long time to come. Though he had moved out of the limelight after he left Voodoo Extreme, he had set up a new endeavor with his friends at Gaming Groove. This site was launched late last year, and appeared destined to become a solid voice in the industry. While details of his death have not been released, I am positive that it was because he was a rabid vegetarian! All joking aside, Billy had a tremendous impact on the industry, and he will be greatly missed. My first real encounter with Billy was when I posted a terrible article about KNI vs. 3DNow! Billy was very gracious and posted the link on VE, and even said some nice things about it (though my article was absolute, complete garbage). I had other interactions with the man, and these were mainly late night emails that talked about gaming and video cards. One that sticks out in my mind was when I first started playing System Shock 2 on a Matrox G400, and all Billy had to say about that was "The G400 is a good card, and SS2 is too creepy for me!" This all seems very odd to me, as I had chatted with Billy not that long ago. In fact Kyle Bennett and I were just talking about Billy and the impact he had on our respective sites, and how his linkages could certainly make or break a website. It was surely an awesome power at the time, and he often wielded it with great mercy and compassion. You will be missed Billy. I hope you have found some peace where you are going. Billy was only 33. He is survived by his son Dax, his ex-wife Aimee, his mom, step dad, and his three little brothers. March 14, 2005 Monday Morning and other Unpleasantries - Josh * Edit: One thing that also rears its ugly head into the mess is Intel and AMD's broad cross-licensing agreements. Would such agreements encompass front side busses and cHTT? I guess theoretically AMD could have used the P6 GTL bus for the Athlon, but at the time EV-6 was much more advanced. Again, AMD probably could have used the GTL+ bus for the Athlon 64, but that would have precluded the "glueless SMP" support that HTT gives. Most likely Intel would have rather AMD taken that route, because then any AMD Athlon based chipset would require the royalty fee to go towards Intel. I mean, who wouldn't jump at the thought of every time your competitor sells a product, you get $5 from it? Still, if the cross-licensing agreement covers things like cHTT, then why is Intel developing their own? This decision still doesn't make much sense to me.* I have been reading quite a few responses to my HyperTransport mini-editorial, and I have come to a few more conclusions. First off, it appears as though the cache coherency HTT protocol that AMD uses for its CPU's is different from the regular HTT links between other chips. While the standard HTT links are royalty free for those that join the consortium, apparently the cHTT that AMD uses is not. Now, this brings up some sticky situations. Why in fact would Intel join the HTT consortium if it can't utilize cHTT without paying AMD royalties. The next bit of nastiness would be if Intel did decide to join, create a cache coherency HTT link of its own, and then be forced to share that technology with the other partners. Now, why does AMD have the power to not share its own cHTT in this consortium? Part of the contract that companies sign when joining the consortium is that any modifications or advancements of HTT should be freely shared with the rest of the group. Apparently the cHTT work that AMD has done was before they started the consortium, and perhaps it is Grandfathered in? I am not sure on the details, but I am trying to contact some folks at the HT Consortium to get a better understanding. The next issue that would be a giant stickler for Intel is that they make a tidy profit off of licensing out their Front Side Bus to guys like ATI, NVIDIA, VIA, SiS, and ULI. Not only that, but due to the royalty charge Intel can have a lot more freedom in pricing their own chipsets. So, if Intel decided to create a Northbridge/Processor HTT bus of their own, would they still be able to charge these other companies royalties? I guess that all depends on how the Consortium really works in this case. If Intel can get the same protection that AMD's cHTT gets, then it can still hold onto the technology and charge its royalties to the other chipset companies. Not only that, but because the Intel and AMD architectures are so different, I find it highly doubtful that they both could eventually fit in the same socket (eg. use an AMD processor on an Intel based motherboard). The biggest hurdle to that would be the integrated memory controller the AMD processor has, and the chipset based memory controller that Intel has. So, even if Intel did join the HT Consortium, and developed a HTT link that worked with the Pentium 4/Pentium M/Xeon processors, we will still have different socket types, as well as other basic changes to the link. So, theoretically, if the HT Consortium gave Intel the same freedom as AMD enjoys, then they could still enjoy the advantages that HTT brings, but not have to change their business model. The more I look at it, the less enticing it is for Intel to join the HT Consortium. The technology is good, but there could be too many legal loopholes that would give Intel the short end of the stick if they joined. And this is mainly because of the way Intel has set up its chipset business model (which provides Intel with a healthy profit each quarter). Now, I am a big fan of open standards, and I need to research this a bit more before I can really come to a solid conclusion. Needless to say that on the surface it appears pretty cut and dried, but once the prying begins then cracks appear everywhere. Since AMD doesn't want to be in the chipset business, and they also want to spread their processors around the market in as large of numbers as possible, they don't charge a royalty to chipset makers to utilize the cHTT bus to communicate between the processor and the chipset. Now, if Intel/IBM/VIA or any other processor company wants to use cHTT in their processors... that is where the problems begin. With this in mind, it is a bit more understandable that Intel would want to protect its interests in this situation. They do not want to give AMD any money whatsoever, which is why they are designing their own point to point link (iHTT?- haha). So while this link's functionality will be nearly identical to what HTT does, they won't have to pay AMD a dime to use it (they have basically taken the idea and engineered it themselves). Though with all this thinking, I harken back to the day when Intel and AMD did in fact share the same busses (namely Socket 5 and Socket 7). At that time there were no royalties to pay, and Intel still made the best chipsets of the time (many AMD enthusiasts used the 430 HX and TX based boards for their K6 processors). VIA, SiS, and ALI (now ULI) all had somewhat buggy products out, at least compared to the Intel chipsets. I think I see the balancing act that Intel had to go through with the licensing of their FSB. The big question they probably faced was if they could make more money and starve their chipset competitors out by licensing the bus and not allowing any AMD products to use Intel chipsets? Or did this decision give their chipset competitors a "sheltered bay" where Intel couldn't compete with them (eg. the Athlon market where Intel is not competing with them on the chipset level)? I think in the end Intel made a mistake here. I think that their chipset products could have pushed its competitors even farther if they had been available for AMD platforms, and in the years of the Athlon and Athlon XP, Intel could have still easily held onto the processor market even though AMD processors could run on Intel based motherboards. Today would be a different situation though with the outstanding Athlon 64 products. Anyway, that last bit was all speculation of course, and with Intel and AMD diverging so significantly for the past two years it is all been rendered moot. However, in 2007, we will see Intel release these "AMD-like" technologies- namely their CSI interconnect and integrated memory controller. Admittedly, now that processors are so complex, their respective busses they use to connect to other components can't really be cookie-cut out and used for all products. The needs of the Pentium 4/Pentium M are significantly different than the AMD Athlon 64. Design choices that were made years ago have pretty much precluded any common bus at this time. We shall see what the future holds though, but I have the feeling that the chipset and motherboard makers would love to see a common link between Intel and AMD products, with the only physical differences be the socket in which the processors were placed. That way they would only have to create one set of SKU's to address both AMD and Intel processors. I know, I know... wishful thinking. This HyperTransport situation is about as clear as mud. NVIDIA Ships Over a Million NF4 Chipsets - Josh This was announced a bit ago, but only now have I had the chance to really weigh in on it. When the Athlon 64 was first introduced, VIA had the defacto chipset for that architecture. The K8T800 series had all the features, as well as slightly greater speed, than anything else out at the time. It was included in most of the first generation motherboards for the Athlon 64 (excluding the Opteron based boards of course, which used an AMD designed chipset). NVIDIA, who had a fairly successful AMD part for the Athlon XP market, introduced the nForce 3 150 series. These fared well in the new Athlon 64 market, but due to some limitations (eg. not a full HTT bus) it was not as widely accepted as the VIA product. This of course changed with the release of the nForce 3 250 products. Finally NVIDIA came out with a product that matched the VIA solution, and in fact was superior due to the high functionality and better overall performance (gig-e, true integration of SATA, etc.). Some 4 months after the introduction of the nForce 3, NVIDIA announced that they had captured more than 40% of the AMD market with their chipsets. Now that the nForce 4 is out in great numbers, this market share should increase again. The nForce 4 is the most common AMD/PCI-E based chipset in that market, while the new VIA K8T890 series is struggling to get off the ground. The main reason for VIA's problems is that the new VT 8151 southbridge is still not available. This has made many motherboard manufacturers shy away from the VIA product due to their somewhat lacking feature set as compared to NVIDIA's nForce 4. Throw in the SLI support, and NVIDIA has a pretty well rounded package from the low end to the high end. With nForce 4 Ultra based boards hitting the $99 price range, it is no wonder that this chipset is now dominating the AMD/PCI-E space. Now that nForce 4 has made it into the Intel space, this should mean that NVIDIA will start to make some pretty good scratch from its chipset division. NVIDIA has certainly gotten itself back on track, especially considering the unpleasantness that it experienced from 2002 til mid 2004. Now it has a rock solid and fast architecture for its graphics (GeForce 6x00 series), as well as outstanding chipset products (nForce 3 250, nForce 4, and nForce 4 for Intel). They are certainly hitting on all cylinders at this time. On the ATI side, they are now offering their Radeon Xpress 200 series of chips for both Intel and AMD. These also appear to be very strong products with plenty of features. Perhaps not as tightly packed as the NVIDIA products, but considering how long ATI has been attempting to enter this market, I would say they are doing very well. Not many motherboard manufacturers have embraced the ATI products as of yet, but I expect to see more go that route. The biggest attraction to the Xpress 200 is the integrated graphics functionality. Their integrated graphics are hands down the best in the industry, and with so little integrated graphics support on the AMD side (currently VIA and SiS are the only competitors that really offer integrated graphics for the AMD Athlon 64) I would expect the Xpress 200 to become very popular with OEM's. This is one area that NVIDIA hasn't addressed since the nForce 2 and its GeForce MX based graphics. Many wonder when NVIDIA will in fact integrate the 6200 series into their chipsets, but no real timetable has been released on this. So, NVIDIA has some very healthy competition from ATI in this market as well. Now all ATI has to do is actually build their marketshare here (which is pretty low at the moment). March 11, 2005 Intel Creating Own "HyperTransport" - Josh Several years back AMD and a few other IC companies helped to found the HyperTransport Consortium, and it has turned into a very viable part of the industry. AMD has shipped millions of HT enabled products, as well as Broadcom, NVIDIA, Agilent, Alliance, and many others. It is a very solid technology, and its uses are pretty far flung (at least within the computer market...). The current HyperTransport specification is up to 22.4 GB/sec bi-directional data bandwidth, and all of this is offered in a very low latency package. This is a lot of bandwidth, and the flexibility of the spec is also pretty tremendous. Plans to scale performance from 1 GHz to 1.2 GHz and 1.4 GHz are in the making. The jump to 30+ GB/sec is not far away, though I am unsure when exactly this will be implemented. For the longest time Intel has been poo-pooing HyperTransport, saying that it doesn't really fit the needs of the industry, and that it really is second rate as compared to the PCI-E architecture. People outside of Intel were somewhat confused by this response, as the two technologies are really complimentary as compared to being competitors. PCI-E is an excellent next generation general bus technology, and AMD and many others are a member of that architectural group. HyperTransport on the other hand excels in point to point topographies and applications. The technology is mature, and very well supported. Well, it now comes to light why Intel never wanted to join the HyperTransport Consortium. They are creating their own nearly identical standard. Intel is creating their own CSI standard to connect next generation Itaniums and Xeons together in multiprocessing environments, as well as utilize this technology to connect their processors to their northbridges and southbridges. This has left many people scratching their heads, as it makes very little sense for Intel to "re-invent the wheel" in this circumstance. There is already a well supported industry standard that has proven itself to be highly efficient and fast in multiprocessor environments (the Opteron in this case), so why waste millions designing your own specification that will most likely not significantly outperform what is already available? Well, two reasons come to my mind. The first is that if Intel embraces HyperTransport and joins the consortium it will validate AMD's efforts and essentially show the market that "AMD got this one right". As bitter competitors, this of course will never happen. The second reason looks to be "NIH", or "Not Invented Here". This attitude has squashed so much innovation in the industry, and it has turned neutral companies into bitter enemies. Now, Intel will probably justify their course of action with something like, "Well, HT is good, but it doesn't adequately address our future needs and lacks some of the very necessary features we want to support." (Ed. Damn, I should go into PR with comments like that) As far as I am concerned, that is a very backwards attitude. This is the reason that HT is a consortium, it has many different companies coming together, sharing what their needs are, and helping to hammer out the specification. If Intel had joined the consortium, then they could have pushed these "features" and probably have them integrated into the specification (honestly, a company of the size and stature of Intel is going to get their way in a consortium of this type). But, I think that Intel sees this as "aiding the enemy" in many ways, because these features could then be implemented into their competitor's products. Still, how many millions is Intel spending to design, test, and implement this technology? If I were a shareholder of Intel, I would raise a bit of a ruckus (especially if I were a major shareholder). How long will this technology take to bring to market, how long will it take to implement into the Itanium and Xeon architectures? If Intel had joined the HT consortium years ago, perhaps we would see a 1.2 GHz HyperTransport bus that stretches between the Pentium 4 and its northbridge? Considering the issues that Intel has had with their 1066 MHz FSB, one would think that HT would be a perfect fit for their problems (especially considering how AMD has no problems with their 1 GHz-8 GB/sec HT link). So, the long and short of this matter is that Intel is spending millions in R&D on this technology, which it really needn't do. HT is a very solid industry standard, and any input from Intel would have increased the performance of features of the standard. Instead, Intel appears to be taking a NIH attitude and is working hard to re-invent the wheel. I know that sounds like a harsh assessment, but in this particular case Intel's attitude has hurt themselves and its customers. If Intel had embraced HT, then they could probably already have HT enabled products, which would have increased the efficiency of Intel's desktop and server products. To this outsider, it just seems pretty silly. March 10, 2005 News, News, and More News - Josh You can definitely tell it is March now, as the industry is really picking up some steam. I always hate the time between the end of Christmas and the end of February, as it seems the industry takes a nice little nap. Now we are overflowing with news! AMD has released their Turion 64 products, which are the new mobile Athlon 64 parts. These are based on the latest E stepping of the Athlon core on the 90 nm/SOI/Strained process. The maximum speed at which these parts run is 2 GHz with 1 MB of L2. Now, the kicker here is that they have a max power rating of 35 W TDP. In fact, there are two types of these chips, the 35 Watt, and the 25 Watt. This should allow the Turion 64's to go into smaller, lighter notebooks and tablets. The Athlon 64 compares well to Intel's Celeron in overall performance, and may have a slight edge in gaming type applications. From the looks of things, I would say that AMD's 90 nm process is a huge success, as they are quickly transitioning all of their lines to that process (the poor Athlon XP and Socket A Semprons are soon to bite the dust altogether). While AMD doesn't offer the same type of package with the Turion as Intel does with the Centrino, they are building some momentum for themselves with a very solid product and a excellent of partners to help them along (eg. ATI, VIA, NVIDIA, SiS, etc.). Anand has a quick preview of the products and technology. I think AMD will really push the features onto notebooks to try to knock down the lead of Intel a bit (features such as Gig-E connectivity, 802.11 G wireless as the base standard, higher performance 3D graphics, PCI-E, mobile SATA). NVIDIA made a couple of announcements here at the beginning of CEBIT in Hannover that should affect a couple of people. The first is that they are now officially introducing the 512 MB GeForce 6800 Ultras. Now, Gainward jumped the gun here about a week or two back, but everyone knew that 512 MB cards were eventually going to hit the market. Now, it is official. While NVIDIA doesn't mention this in their press release, it has been said that the higher density GDDR-3 is still not out in volume yet, so these 512 MB boards contain 16 memory chips in total (as compared to the usual 4 or 8 chips). This of course necessitates a PCB respin for these initial products to accommodate the memory. Many expect good supplies of the higher density GDDR-3 products in about a month and a half. That of course coincides nicely with the upcoming Spring refresh for ATI and NVIDIA. The second announcement that NVIDIA made is that they will be giving another preview of the nForce for Intel, and the SLI version will be running 2 x 512 MB GeForce 6800 Ultras. That's 1 GB of video memory in one machine. Hell, most of us are still "just" using 1 GB of main memory in our machines! NVIDIA is also showing off their native 6200 AGP product at CEBIT, and this looks to totally wipe out the GeForce 4 MX and GeForce FX 5200 series of products. This is something that many people have complained about for years with NVIDIA, and now we are finally able to get solid DX9 products from top to bottom. After looking at the GeForce 6200 TC's core, I come away with the overall impression that it is really tiny. This should be a very good seller for NVIDIA, as the boards and chips are very cheap to produce, and will make a solid impact on the retail side of the market. So, those of you out there with a GeForce 4 MX or GeForce FX 5800 or below, this is going to be one very cheap upgrade for you. Speaking of SLI, apparently NVIDIA is sending me a pair of 6800 GT's to start tearing into. I have had some interesting requests in regards to SLI as of late, so I will be giving these puppies a really good workout! I also have the Albatron GeForce 6200 TC 128/256 MB version (which I believe actually has 64 MB of onboard memory, and uses Turbo Cache to reserve the other 64/192 MB in main memory). I am actually quite interested in the possibilities of TC, and how a couple of system tweaks could really make or break the performance of that architecture. Needless to say, I am quite enthused about both SLI and TC. Chaintech is also sending me their VNF4/Ultra motherboard, which looks to be a very solid and cost efficient nForce 4 Ultra board. These can be had for slightly less than $100, and it packs in all the features of the nForce 4 Ultra. Should be a very interesting review! I also have the MSI K8N Neo 4 Platinum, and am expecting the new Tyan Tomcat K8E in very shortly. Not to sound like too much of a fanboy, but NVIDIA has certainly been on their game since the release of the GeForce 6800. In that time they have had the nForce 3 250 series, the nForce 4, the nForce 4 for Intel, the 6600 series, and the 6200 series. Not a bad group of products to be released in less than a year. And apparently, it will continue to roll out products at this rate for some time to come. ATI is doing their best to keep up, and they also have a lot of good products on the market, but the one area that is of great concern to a lot of people is their support of the AGP platform. While the transition from AGP to PCI-E is something to concentrate on, ATI continues to support the AGP platform (especially in the budget and midrange markets) with older products. Now that they have released their Rialto PCI-E to AGP bridge chip we will see a refresh for these markets, as AGP is still very important on the retail side. A retail badged X700 AGP would be a very solid product for ATI's partners, as it costs less to produce than the Radeon 9800 series of cards, and it would compete very well with the GeForce 6600 GT for AGP. ATI has released the latest Catalyst 5.3 drivers for your consumption, and these improve on the series with some tangible advancements. The biggest news is the new overclocking/Overdrive functionality for the cards that support it. It should allow the user a lot more flexibility than the previous Overdrive tab, as they can set both core speed and memory speed. Quite a few bug fixes in this one, but there still appears to be quite a few unfixed bugs as well (according to their readme). You can get the non-CCC version here. 3D Chipset is reporting that they have a copy of the latest Beta Forceware 76.10 drivers. These are the latest to hit the web for NVIDIA, and it apparently has a couple of bugfixes of its own. Now, I am all for leaked drivers, but I honestly wish that NVIDIA would have an official Catalyst Crew of their own that would release an official driver once a month. Apparently NVIDIA is aiming for two official driver releases a year, and I think that kinda pushes things a bit. NVIDIA is releasing more official "Beta" drivers these days, but I think many users and OEM's would prefer an official once every month or two type release of drivers. More news is being put out about the X-Box 2, and it is looking to be a very impressive piece of hardware. Currently Microsoft is sending out developer kits with dual PowerPC processors and the R500 video card. Now, the R500 will not see release into the PC market (just like the NVIDIA NV2A which powered the original X-Box), but the very fact that it exists means that ATI has made their first 90 nm product. The R500 is a complex piece of silicon, and to have working versions at this point speaks volumes about ATI's ability to bring a 90 nm product to market. The silicon should still be considered beta, but at least it is working and is feature complete (or so they say). This probably means that the R520, which is the Spring refresh product for ATI, should also be well on its way towards delivery. I have not heard if ATI has any of these chips back, but I think we can safely assume that it taped out a while ago and they are probably expecting first silicon very soon. I keep saying that this Spring will be interesting! March 9, 2004 Clarifications on DDR/DDR-3 and News of Interest - Josh I have had a couple of interesting comments so far on the AMD/DDR subject, so I thought I would relay them here. DDR-3 and GDDR-3 are not identical products, but we can certainly see what the transition in design from GDDR-2 to GDDR-3 entailed. Better termination, better power draw, faster overall speed, and a reworking of some of the basic signaling techniques from the memory controller to the memory itself. If AMD does decide to go with DDR-2, they will have to introduce a new socket infrastructure to accomodate the changes. Socket M2 is AMD's 1207 pin socket that will be introduced in 2006 with a series of new Athlon 64 cores. This will allow the use of DDR-2 or DDR-3 (whichever product AMD decides to choose), as well as enable other connectivity features for the Athlon 64. Am I saying that DDR-2 is bad? Not really. DDR-2 has a lot more potential, and when it hits DDR-2 800 speeds, then it will be a lot more competitive with DDR 400 in terms of overall latency (with double the bandwidth). Perhaps AMD is merely waiting for DDR-2 800 before it jumps into the water. By early 2006, there will be a lot more support for DDR-2 due to the juggernaut of Intel pushing the standard. If the speed is up there, and AMD thinks that the latency penalties are certainly overcome by the high speed, then perhaps AMD will in fact support DDR-2. I still have my doubts whether they will or not, mainly due to their lukewarm reception of the technology to date. I think they are hoping something more palatable will come out before then, and are working with their memory partners to make it possible. Perhaps we are looking at this the wrong way though. Perhaps the reason that AMD is not excited about other exotic memory types is simply price. When AMD was pushing DDR memory, the only other alternative was RDRAM at the time. SDRAM was still king, and it was inexpensive to produce. DDR was a great fit for AMD as it doubled bandwidth with only a 5% premium in manufacturing over standard SDRAM. RDRAM was prohibitively expensive, and the serial design left much to be desired in terms of latency. Perhaps we are looking at the same situation? Many have suggested that AMD is in fact looking at other options such as FSRAM or MRAM, or even perhaps RAMBUS's XDR (aka Yellowstone- which will be used in the PS3). The crystal ball here is rather fuzzy, and the only people who have a better idea than most of us are the engineers working on the next gen Athlon 64 parts at AMD. I think their overall decision will probably be based on what is easiest to produce, cheapest for the market, and performs adequately. Perhaps I just essentially described DDR-3... but perhaps not. The memory market can be very liquid and choatic in times of major transition, and often the major players in the arena do a lot of political maneuvering to push the standard that they want (examples of products that were pushed by a select few but eventually came to nothing include RDRAM, VC-SDRAM, etc.). If DDR-3 follows many of the design changes that GDDR-3 went with, then it would actually be a simplified design of DDR-2, and run at higher bandwidth rates. Latency with DDR-3 should not be significantly higher than DDR-2, and the overall increase in speed will be an added benefit. Again, the specification for DDR-3 is not set in stone, and we could in fact see a product that is significantly different than GDDR-3. Once a final specification comes out, we will be able to more adequately speculate on which way Intel and AMD will go with their underlying memory support. The only certainty I have now of the situation is that AMD will not adopt DDR-2 in 2005. After that... it is anyone's guess. AMD Close to 4200 Release? - Josh Some leaked screenshots of what appears to be the Athlon 64 4200+ processor have found themselves on the web. This appears to be the San Diego core (E0 stepping, 90 nm, SSE-3 support, revised memory controller) running at 2.6 GHz. No word was mentioned on how high this chip could possibly clock, but from the indications that the previous Winchester core gave us, 2.8 GHz should be easily attainable. I hate to mention that 3.0 GHz should also be something they can get to, but the last time I suggested that Damon Muzny of AMD just laughed at me. Not sure what he really meant by that, but my impression is that he doesn't think it will be that easy to attain. What is not known is if the leaks are coming from a reviewer with the product in hand, or a Taiwanese employee of some motherboard manufacturer who has their hot little hands on a sample that AMD sent for motherboard validation. Now, we know that AMD is actively selling Opteron 252 products (which are also based on the E0 revision at 90 nm), and AMD's 90 nm process is very clean at this time. So, it is pretty apparent that AMD could probably introduce the 4200 (and most likely the FX-57) within the next two months. There are those that are pointing towards the end of March as a good time, as AMD again cut down the prices of some of their high end processors (the 3800+ is under $380, and the 4000+ is under $560). This usually means that AMD is preparing the market for some new high end SKU's that will fit in the $600+ range. Not only that, but supplies of FX-55 chips are quickly dwindling on many retailers' sites. With that in mind I don't see a reason why AMD won't be introducing something in late March/early April. It was October 2004 by the way when the FX-55 and Athlon 64 4000+ were released. So the time appears ripe for something new. Ageia PhysX PPU - Josh In what looks to be a fascinating product, a small company named Ageia have invented a new niche for themselves. They are introducing what is the world's first dedicated Physics Processing Unit. This 125 million transistor chip promises to revolutionize the gaming industry, as well as any software application that utilizes advanced physics computations. While this is mainly aimed at gaming, I can see some enterprising software people utilizing the specialized horsepower of this product to accelerate other applications. While CPU's and GPU's are very good at what they do, the people at Ageia decided that with all the separate functions that each of those products have to perform, there was definitely room for a PPU. GPU's have offloaded a lot of functionality from the CPU, but game designers have increased the general CPU usage by performing higher function AI routines, as well as other "features" that can bog down a CPU. Graphics chips, while taking more and more computing responsibility for themselves, are still not able to do it all. A dedicated PPU in a machine can have a tremendous impact on future titles that utilize this technology. While a modern CPU can adequately handle a good amount of physics based interactions, the use of this specialized hardware can have a 1000 fold performance advantage over a CPU alone. Many are likening Ageia to 3dfx (well, 3Dfx when it was first introduced into the public eye) when they released their Voodoo Graphics. While there were other competitors in the 3D arena (NVIDIA, Rendition, Matrox, ATI, etc.) the Voodoo Graphics revolutionized the 3D industry with a solid design that did all it was supposed to do at a reasonable price (first introduced in the $300+ range). If there is a killer app being developed to utilize the PPU technology, we could see other competitors join this field (definitely something both NVIDIA and ATI are looking at with great interest). As it is, Ageia is out first with their product, and they expect to have a retail presence by the Holiday Season this year. This technology could be a blessing for game developers who really want to wow their audiences with stunning graphics, advanced AI, and a physics system that would knock your socks off. I am keeping track of this with great interest. The product itself would be a PCI or PCI-E card, and the initial test board has 128 MB of memory on it. The same board itself can be used in either a PCI or PCI-E socket, so there is no reason to have two SKU's for this product. It only eats up about 25 watts so far, and for a 125 million transistor chip... that ain't bad. Obviously it is not running at a high core clock, and it is also a highly parallel design (so it doesn't really need to run at high clock speeds). AMD's 1 MB Athlon 64 processors are also around 125 million transistors, but much of that is taken up by cache. Those are about the only details that Ageia is releasing on this product, but I imagine we will hear a lot more about this technology the closer we get to the end of 2005. I can foresee motherboard manufacturers integrating this part onto their higher end boards and marketing these as the "total gaming hardware solution". This certainly adds an interesting wrinkle to things! March 8, 2005 AMD, DDR, DDR-2, DDR-3, and PC4000 - Josh There has been a lot of talk lately about AMD's possible move to DDR-2, so I figured I would throw my $0.02 in. DDR-2 is an interesting technology, and it has some features that really make a lot of sense. I suppose the most important for many is the on-die termination. This small change to the basic DDR design makes it significantly easier and cheaper to implement DDR-2 on a design. If you take a look at any DDR motherboard, you see a whole line of resistors stretching from end to end of the DIMM slots. These are the terminators for DDR. Now, with on-die termination, there is no need to place those resistors in a design. MS over at Lost Circuits has done some exhaustive articles relating to DDR-2, and the very simple fact of the matter is that clock per clock DDR-2 is about 10% less efficient than DDR. Not only that, but currently it is more expensive to produce than DDR. The only real other advantage that DDR-2 has over DDR is that it runs at 1.8 volts (vs. 2.6 for DDR). While it runs at lower voltage, it still pulls a lot of power, and heat is definitely an issue with these products. So why is Intel using DDR-2? I honestly have no idea. The increased bandwidth is certainly nice, but the latency penalties to the design are certainly a significant downside. DDR-2 won't be competitive with DDR in most applications until it hits 667 MHz and above (800 seems to be the magic number where its overall performance goes beyond regular PC3200 DDR). The next question is if AMD should eventually go with DDR-2? I think this really depends on two factors. The first being how well AMD sees its memory partners producing DDR-2, and how popular DDR-2 DIMMS become with OEM's. Samsung is making a very strong push with DDR-2, but many of the other DRAM folks are not making much of an effort in that direction. Samsung certainly made a strong impression with their TCCD based DDR DIMMS, but with the internal pressure to push DDR-2, they have apparently cut down production of TCCD and really only Corsair has a contract in effect that Samsung would supply them with as many TCCD IC's as they want for the next 6 months. This was obviously a strong move by Corsair, as within the next few months they look to be the only DIMM manufacturer that will have TCCD based products (which are in very high demand due to their excellent stock performance, as well as the ability to clock above 250 MHz with fairly tight timings). So does this mean that there will no longer be high performance DDR DIMMS from any other manufacturer? Of course not. Apparently Winbond is getting back into the act with an improved version of their legendary BH-5 chip. Micron also has some high performance DDR IC's that they could also sell to other 3rd party manufacturers. There are definitely options out there, and some company will eventually step up to the plate with a product that will replace the TCCD DIMMS in the general market. The second factor would be how fast DDR-3 will hit the market. DDR-3 gets rid of many of the problems with DDR-2, and in the graphics realm GDDR-3 has been an overnight success due to its high efficiency, lower power consumption, and high clock speed ceiling. DDR-3 for mainstream applications would bring many of the advantages of GDDR-3 to the processor market, as well as provide for greater overall performance. While the exact specifications of DDR-3 are not yet ratified, I think that AMD is working very hard behind the scenes to get the technology to the market as soon as possible. While many of the features of DDR-2 are very good, the drawbacks and compromises to the overall design make it a very unpalatable product for AMD. Since AMD put the memory controller on their CPU, cutting down on latency was a huge factor for that decision. Why would AMD want to introduce significantly more latency to their design by adopting DDR-2? My gut instinct is that we will never see DDR-2 actually implemented into the Athlon 64. So where does this leave AMD? Well, so far PC3200 DDR SDRAM in dual channel mode on the Athlon 64 is keeping up with the processor speed increases. While any CPU likes more bandwidth and faster data requests, PC3200 will serve AMD's needs for at least the next 9 months. Dual core processors will of course need more memory bandwidth, and because of that we might see a push by AMD and many memory manufacturers to make DDR 500 an actual JEDEC specification (DDR running at 250 MHz, or PC4000). Most major DIMM manufacturers already have "official" DIMMS marked at such speed, so it may be just a short time before that specification is ratified. This is not hard to imagine, especially since the memory controller on the latest Athlon 64's can run memory at 250 MHz with no problems (and OCZ has shown their memory running at DDR 772 on the Athlon 64- that's 386 MHz for those with problems doing division). Running at 250 MHz at 2.5:3:3:8 timings actually shows better single and multiple data access performance than 200 MHz DDR at 2:2:2:5, not to mention the bandwidth increase (which is a big factor in any streaming application- something that DDR-2 actually excels at). So, I would expect an official PC4000 spec for DDR before we see AMD support DDR-2. If you remember back in the day when Intel was all over RAMBUS and AMD was the driving force for DDR. Many scoffed at AMD, but in the end they were proven right. DDR was the right technology at the right time, and now it is the dominant memory type for all PC markets. It appears as though AMD wants to skip DDR-2 altogether. While the idea of DDR-2 was good, its implementation left much to be desired. I think that AMD saw that early on, and while Intel has really pushed DDR-2, I think it is going to be a dead end. DDR-3 is probably going to be the next big thing, and many expect to see the first implementations by the end of 2005. DDR-3, even though the specification isn't official yet, looks to build on the positive aspects of DDR and DDR-2, and should have higher efficiency than both of its predecessors. It will also have a simplified design, and from initial indications will be very efficient in terms of power consumption and heat production. These are all very positive aspects, with very few negative ones. I could be very wrong with my assessment, but after looking around and talking to some people in the know, this is the conclusion I have come to. There is no real reason for AMD to adopt DDR-2 into their Athlon 64 design, and DDR will continue to be the memory workhorse of the industry throughout 2005. It still has very good bandwidth in dual channel architectures, and its low latency design makes single and mutliple d-word read/writes very fast and efficient (something that DDR-2 cannot boast of). I think we will start to hear very soon whether AMD and other memory companies will push for an official PC4000 specification, and when that happens we will have a much better insight into what AMD is planning. A Athlon 64 on a 4 x 250 MHz HTT bus (1 GHz HTT) running dual channel PC4000 at 2.5:3:3:8 latencies is a very fast and efficient setup, and essentially matches anything that Intel has in its arsenal (namely the i925X products). It would be a very good step, especially considering many overclockers consider that a very easy target to reach with their latest motherboards, processors, and memory. So will we see this implemented in the upcoming dual core designs? Hard to say... but seeing that the first dual core chips will be in the Opteron product line, I don't thing we will see PC4000 in that space. Once dual core hits the desktop though... that could be a different story. PC4000 here we come! March 7, 2005 Well, that weekend certainly flew by! I was actually quite busy with two reviews, the MSI K8N Neo 4 Platinum (nForce 4 Ultra) and the Sapphire X800 XL. So far I have been quite impressed by both of these products! The MSI boards runs so much better than my previous 939 testbed (which was the Gigabyte GA-K8NS-939 Ultra- a very unimpressive product). This board runs at 250 MHz HTT with no qualms, which takes my Athlon 64 3200+ Winchester up to 2.5 GHz on stock cooling with a small voltage increase. There still seems to be quite a bit more headroom for both the board and the chip, but I am going to have to more thoroughly explore their abilities. The Sapphire X800 XL is a very interesting product, and if a user can actually pick one up for $299, it is a very solid investment. The X800 Pro is about $100 more expensive than the XL's MSRP, but in most applications it is slower. The only area where the X800 Pro takes any kind of lead are in vertex shader heavy programs. Of course, the only real VS heavy programs I know of are the benchmarks that specifically test that functionality (namely the 3D Marks). In every other test I have run, the X800 XL literally walks all over the X800 Pro. Things of course even up when the X800 Pro is clocked to 550 MHz core, but at stock speeds the X800 XL is significantly faster. I am very perplexed about ATI's product placement, as the X800 Pro for AGP and PCI-E just do not make sense in the face of the X800 XL. The XL chip is cheaper to make due to its 110 nm process, and yields on that process are supposed to be much better than the 130 nm Low-K process. Next is that the X800 XL features 256 MB of GDDR-3 that runs at around 496 MHz (992 MHz DDR). The X800 Pro's memory in comparison runs at around 450 MHz. I just cannot understand why there are any X800 Pro's on the market anymore, especially the new PCI-E based models. In nearly every scenario, the more expensive model runs slower than the XL. Now that ATI has the Rialto bridge chip running, we are starting to see the X800 XL's being ported over to AGP. Now, online retailers know this, which is why they charge around $375 for the X800 XL, and around $395 for the X800 Pro. These guys are doing their best to milk as much money from users (and other resellers) as possible for this product. When looking at the performance of other parts in this same category, I guess they are somewhat justified. What is a bit more confusing to me is that many of ATI's partners are offering X850 Pro's to the market. These run at 500 MHz core and 1050 MHz memory, and I would imagine that these give the X800 XL a bit more of a run for its money. Still, the X800 XL will have around 400 MPPS fillrate advantage over the X850 Pro, but the memory is of course slower. Now, I have not yet tested the ability to overclock the X800 XL as of yet, but with some judicious fan control, I would imagine that 450 MHz core is not out of the question, as well as nearly 1100 MHz with the memory (it is 2.0 ns GDDR-3 by the way). So, depending on how the rest of my testing goes, the X800 XL could be one of the most exciting mainstream products to hit the market in a long time. It keeps up with the GeForce 6800's of the world, and with some tweaking it can be an exceptionally fast card. The standard X800 (if it ever reaches its $249 price point) looks to be another impressive product, especially at that price point. While it doesn't have the clockspeed of the X800 Pro (400 Mhz vs 475 MHz), it can most likely be clocked to around 440 MHz and slightly above. For the price they are offering it at, I would happily go with a X800 over a similarly priced GeForce 6600 GT (especially since the X800 features 256 MB of memory on a 256 bit bus). Unfortunately for us though, the X800's are still priced around $299. One thing I should mention though, big stores like Fry's and Best Buy always go with the MSRP, so while these products are priced at $299 (and the X800 XL is at $375) online, the big stores are cutting users a big break by hitting the MSRP! This is quite a change from the norm, where the brick and mortar shops typically charge more for products then their virtual and online brethren. Now, I am actually quite impressed by ATI at this time due to the excellent availability of ALL of their products. You can purchase the X800 XT PE AGP for slightly above $500 at some places, and even the latest and greatest X850 XT PE's are available in good quantities (though at $590 to $650, depending on the manufacturer). The entire line is well represented in retail, as well as OEM. ATI certainly got their business together in a big way, and this is certainly good for the consumer! NVIDIA is in a bit of a strange place at this time with their products. The supply of high end PCI-E cards is still pretty thin for NVIDIA (such as the 6800 GT and Ultra models), but the low end to mid-range market is well represented by the GeForce 6200 and GeForce 6600 series of cards. The 6600 GT has become the darling of mid-range enthusiasts everywhere, and it is pretty rare to see people in forums (or other reviews for that matter) to recommend a X700 Pro over a 6600 GT. Still, while the 6800 GT's and Ultra's are well represented in the AGP world, they are still scarce in the PCI-E space. Now, NVIDIA did quietly release two new chips into the PCI-E market, the NV41 and NV42. The NV41 is the IBM based 12 pixel pipe, 5 vertex shader, PCI-E native chip made on their 130 nm process. The NV42 is essentially the same architecture, but it is made on TSMC's 110 nm process. NVIDIA clock's these chips at the same speed (325 MHz) for the GeForce 6800 PCI-E cards, but unlike the AGP versions of the 6800, these come with 256 MB. Now, very little overclocking testing has been done on these boards, but most of them do run at 350 MHz and above. Also, data on how the NV41 clocks as compared to the NV42 are very scarce, and I would be very interested in finding out which one typically does better. For the $289 and above that retailers are asking for this product, it could be a seriously good investment at this time. If the NV41 and NV42 can reliably hit 420 MHz, then it would be a very good value. We are actually starting to see a pretty hefty erosion of prices on the 6800 series of cards, and that typically foreshadows the introduction of new parts. We are getting awfully close to April, which historically has been the time to introduce new products into the 3D graphics market. Both NVIDIA and ATI will be refreshing their high end products (but I highly doubt they will do anything to their midrange and low end, as well as probably leave the $249 to $349 area untouched by new products). However, $400 and more will see some new products hitting the review sites within the next two months. That is really all I can say for right now, as I am still working on an update to the State of 3D concerning both ATI and NVIDIA. I have lots of good info sitting here, and I really need to get it out soon! March 4, 2005 Corsair Flash Voyager 512 MB Edition Review - Josh Some time ago Corsair sent me a sample of their Flash Voyager series. These memory sticks go from 128 MB all the way to 2 GB, and they feature USB 2.0 connectivity. These very fast, rugged, and small sticks should be a definite consideration for anyone looking for such a product. Here is a quick quote:
You can find the entire review here.
If you have found this article interesting or a great help, please donate to this site.
Copyright 1999-2005 PenStar Systems, LLC. |
|||
