   |
|||
|
|||
|
:
News
: Archives
|
|
|
|
|
|
|||
|
|
|||
|
|
|||
|
June 7, 2007 AMD Pushes Low Power and Barcelona News - Josh Two days ago AMD released their new low power desktop offerings, the BE-2300 and BE-2350 processors. If you haven't noticed by now, AMD is changing the way they are branding their chips. The BE-2300 is a dual core Athlon X2 64 running at 1.9 GHz with a power rating of 45 watts. The BE-2350 is clocked at 2.1 GHz and also with a TDP of 45 watts. These products are supposed to retail for $86 and $91 respectively, which is a pretty good deal when considering what a user gets. These are intended for not only the HTPC market, but also for anyone that wants a cool and quiet desktop that will still hold its own in heavy applications. These chips are essentially Brisbane cores made on AMD's 65 nm process. Now that particular process has been going full bore for around 6 months now, and AMD has constantly been improving it. This has allowed them to cherry pick dies that can run at much lower voltages, and therefore have a lower TDP at the target clockspeed. So instead of the 1.25 to 1.3 volts the Brisbane core usually runs at, these are sailing along at around 1.1 to 1.15 volts. Unlike the previous low power 90 nm X2 cores, these seem to be able to overclock quite well (look to be getting around 3 GHz for some reviewers). This is about the same as the other standard Brisbane cores, and some samples will of course clock better than others. I applaud AMD on their low power offerings, but this is essentially a niche product at a price that is well below what AMD want their ASPs to be. AMD is of course hoping that this will be something more than a niche product, and considering that Dell and HP bought up pretty much every Energy Efficient/SFF X2 that AMD was able to produce last year, perhaps this push will help expand that particular market. Nobody really likes a loud computer, and these chips will help keep things quiet where they need to be. It is good to see AMD's 65 nm process looking this healthy though, as it potentially means that it is at a mature enough spot to support the next generation of products AMD is pushing to the market later this year. Speaking of Barcelona, it now appears that a September release date is more likely than a July one. This is disappointing to many as AMD needs as much time as possible to potentially recapture some of their marketshare (and increase their average selling prices) before Intel is able to release their 45 nm products. At Computex most of the Barcelona samples that are floating around are clocked at a measly 1.6 GHz, which is not terribly impressive at this time. AMD is hoping to release these chips initially at 2.0 to 2.3 GHz, eventually going up to 2.6 GHz in short order. This means that the Phenom desktop parts will likely not show up until almost the end of this year, and the dual core desktop versions will likely be at the beginning of 2008. It is unfortunate for AMD that they were unable to get this out in short order, as by the time their mainstream 65 nm next gen parts are out, Intel will be shipping (in mass quantities) their 45 nm mainstream parts. Depending on how these two products end up performing against each other, AMD could have a really rough year ahead. AMD still says they are on schedule to begin production on their 45 nm desktop parts in the middle of 2008, and if this is in fact the case, then AMD has cut Intel's process advantage to about 6 months. Not only that, but AMD's 45 nm process could potentially be slightly faster and more power efficient than Intel's. The use of immersion litho, SOI, and a host of other features could in fact make it a world class leading process (AMD is working very closely with IBM, Chartered, and others). That bit of speculation is still up in the air, as we are still a year away from seeing products from AMD made on this advanced process. There are a few rays of sunshine for AMD at this time though, and those are mainly from the graphics division. Apparently OEMs are very excited about the HD 2400 and HD 2600 products, and they could be the dominant midrange graphics offerings in the market this Fall with the OEM refresh. While the HD 2900 XT has failed to awe the larger audience, these new midrange chips could really prove to be a shot in the arm for AMD. Not only that, but the 65 nm refresh of the HD 2900 XT looks to be set for introduction as early as mid-August. My gut feeling here is that the 65 nm version will significantly improve upon the 80 nm product released last month, and it will address many of the shortcomings of the current product. This means improved texturing performance, improved anti-aliasing, and the inclusion of UVD. June 5, 2007 Computex Gets Rolling - Josh It seems the yearly monster convention is in full swing, and many of my colleagues are hitting the streets of Taipei looking for new motherboards, video cards, power supplies, memory, and about anything else computer related. Some of the big news so far is the AMD 790 chipset and the Intel X38. These will be compared to each other quite a bit this Fall I am thinking, and the manufacturer support of these products looks to be pretty impressive so far. NVIDIA will be releasing their next generation of chipsets this Fall when the AMD Barcelona class chips finally start hitting the market, but I am wondering what new and exciting features we can expect from that company. It seems with every new release they have added more functionality to their chipsets. On the homefront I am in the middle of testing the HD 2900 XT with the latest Cat. 7.5 drivers. I am also looking at a couple of micro-ATX motherboards from Biostar which should tickle a few fancies. Gigabyte was also kind enough to send me a P35 based motherboard which looks to be quite a performer as well. I am also knee deep in writing a followup article on a topic I raised about 8 years ago, and haven't revisited much since. Hopefully I can get that up shortly, as it is a pretty fascinating topic (and one seemingly mired down in myths and misconceptions). I am also in the midst of planning another surprise, but it does not look to go live until at least early July. More on that later. June 4, 2007 Payola, Advertising, and Awards - Josh Last night Daily Tech released an article covering the topic of payola and online websites. For some years I have heard of some sites that were essentially "for sale" with their reviews and editorial content, but this wasn't something that I had experienced myself. Daily Tech set up a series of false companies and contacted what they considered the top 150 english written tech sites around the world. They found a pretty good portion of the mid-tier sites would accept some form of payola. Their results are interesting, and it should cause some readers to really question who some sites are really looking after. As you can see from my banners, I don't have a major tech company advertising with me. Which is unfortunate, because my wife wants a new kitchen. All kidding aside, I feel that through the past years I have been pretty even handed in my coverage of the industry, but I do admit that in my younger days I had more of an affinity to certain companies. This affinity was not due to money, but rather to the individuals in the PR departments actually treating me with respect and often giving me a hand with advice, introductions to other PR folks from other companies, and honestly just being a good friend. As I have gotten older (and hopefully wiser) I feel that I have balanced my perspective and can now keep my personal relationships with these people, but still be able to take a good look at the products they are offering. Being even handed and honest will pay off in the end, as those people who take their jobs seriously and have ethics will see you for a honest and fair reviewer. This leads me onto my next point. Editorial awards. I have found that I have given out quite a few of these as of late, and I have been debating with myself if that has been the right thing to do. After a lot of review, I find that in these cases they are deserved. I started a new award this past year simply called "A Penstarsys.com Outstanding Value". These are applied to products in which I feel offer a lot to consumers for their price. Examples of this were the Bluegears b-Enspirer, which was one of the first CMI 8788 based soundcards on the market, and being offered for a very fair price for what the user received. The latest GeForce 8600 GTS from Gigabyte also received this award, which currently retails for around $195 and offers a lot of extras and silent cooling to boot. The Auzentech X-Meridian received the Excellence in Engineering award because it was a unique and high quality design, and offered for less money than other competing products which are based on the reference C-Media design. Something to consider is that I rarely receive a real stinker for review. Most fly by night hardware companies usually try to hit up the bigger sites with their products, and rarely come knocking at my door. As such the cross section of hardware that I review is typically from larger hardware outfits and their products do not suck. So, the question is, should I become a lot more picky about who I give these awards to? I still haven't answered that one completely. For the time being I will keep going as I have with the way I review products, and if they fit my criteria for what an excellent design is, or what kind of value they possess, then I will give them the appropriate award. Make no mistake, there are a lot of good products out there. With honest reviewing from many sites, as well as very active forums where users constantly comment on the products they buy, there is a lot of positive pressure on manufacturers to make damn sure that they have a working and good quality product to put out. We rarely see products like the infamous "Rocco 5.1" hit review sites anymore! So in the end is this good or bad? I am not sure. I know that we must keep alert to people trying to pull the wool over our eyes, or who in fact offer content for sale (in one form or another). We still live in a world where "buyer beware" is something we must consider everyday. May 31, 2007 Auzentech X-Prelude Info - Josh Hardware Canucks were able to get their hands on some information about the upcoming X-Fi based soundcard from Auzentech. From the looks of things, it is certainly a mouthwatering product (well, for those who like soundcards). This is the first time Creative has allowed another company to sell a X-Fi based board, and honestly one of the only times that I can remember seeing a card based on a Creative chip not being sold by Creative. The card is packed. It is basically a cross between the X-Meridian and a 64 MB X-Fi board. It features a replaceable OPAMP for the front two channels, though why a user would want to swap out the LM4562NA is beyond me, as that is probably one of the nicest OPAMPS a user can buy. It is unfortunate that users cannot swap out the center/sub channel OPAMP as well. Typically the front three main speakers want to be matched up as best as possible, and using a different OPAMP for the center vs. the front channels can cause some coloring of the sound that users might find annoying. Fortunately Auzentech appears to be using the OPA2134 units, which are very good in their own right. It is great to finally see Creative selling their chips to someone else, as it will help to give the market a much needed push. The price of the card will be upwards of $200+, so it will definitely be an enthusiast level product. Creative will most likely provide the drivers, as Auzentech really does not have a driver/software group of their own (other than basically packaging drivers from C-Media at the moment). I was really impressed by the X-Meridian, and this card looks to be a step above that in many ways. The card is available for pre-order with a street date of June 6. I am unsure if product will be available by that time, but it certainly appears as though the design is well finalized and ready to ship. Now we have to worry about drivers, supporting Dolby Digital 5.1 encoding in a reasonable timeframe, and hopefully getting DTS encoding onboard as well. Those things all rely on the software folks at Creative from what I understand, and considering Creative's track record as of late... May 29, 2007 The UVD Fiasco and Potential Fix - Josh As most of you have probably read, the AMD HD 2900 XT does not in fact feature the UVD (aka Avivo HD) functionality that its 65 nm stablemates possess. This means that the full acceleration of H.264 and VC1 content does not occur in the 2900, only the 2400 and 2600 models (oh, and the rebranded 2300 doesn't feature it either). Tech Report has a nice write-up of the situation. The official line is that the people that buy the HD 2900 XT will already have a high end processor from either Intel or AMD, and the extra offload that UVD provides will not really be noticed. While this may be true about the buyers of the HD 2900 XT, it is still disappointing to be deluged by all this information about how good UVD is, and then to find out that the top of the line product does not support it. This harkens back to the NV40/PureVideo issue, where the functionality was actually broken. In this case the functionality is not broken, but I do believe that AMD/ATI ran into a die size/transistor count problem. UVD is only a smaller part of the bigger picture. To get the basic 3D functionality that ATI was looking for, as well as hitting their performance estimates, I think the transistor budget they initially were aiming for was quickly surpassed. The 320 stream processors (in groups of five) seems to have taken up a lot of that space, as well as the large caches that the HD 2900 XT features. I believe that ATI had to really cut back on other areas to be able to get close to the transistor budget they had set for themselves. Among the casualties appears to be UVD, slightly simplified render back ends that use the shader core to handle some multi-sampling functionality (like downsampling), and the amount of render back ends and texture units. This is only speculation, but from my testing it just seems like the HD 2900 XT is somewhat unfinished, and this latest revelation somewhat supports this. There is a bright side to all of this though. That bright side is the 65 nm refresh we should expect to be seeing later this summer of the high end. 65 nm should allow AMD/ATI to flesh out that part (potentially the HD 2950), as it will allow quite a few more transistors to be put into the same die space. My guess is that the features and functionality that are seemingly missing in the current card will finally see the light of day. It is easy to speculate that UVD will finally be integrated, but what else can we expect? I don't think that it would be too far out of line to consider that AMD/ATI will throw another quad of render back ends into the deal, giving 20 or so (potentially 24, or two quads extra) and improving the multi-sampling portion so it does not rely on the shaders to do some of its work. This means that AA performance will not take a hit like the current HD 2900 XT does. I think that they would also be smart to increase the texturing functionality, especially since texturing now includes more math functionality than the traditional "apply a texture to a polygon". Again, I would expect to see another 4 or 8 of these units. This is not the end of my speculation though. To get the most usage out of those new units, I believe a lot of work is being done to add performance and functionality to the instruction scheduler, to further offload work from the driver side compiler that the HD 2900 XT relies upon. The VLIW (very long instruction word) architecture that AMD/ATI relies upon with the HD 2900 XT is very sensitive to how instructions are fed to it. In sub-optimal conditions we would expect to see only 64 of the 320 scalar units being used. If AMD/ATI are able to beef up the instruction scheduler to more adequately utilize those 320 scalar units, there is no real reason why ATI would have to increase the number of those units. By improving efficiency to a great degree, as well as provide greater fillrate on both pixel and texture ops, they would have an easier time redesigning that chip, mainly because they could leave the caches and shader units alone for the most part. A large portion of the chip could stay "as is", all the while improving those parts I talked about. In the end such changes could have a tremendous effect on the performance of the chip. Using the 65 nm node will increase core clocks, as well as reduce leakage. The card could be clocked faster yet pull less power. We would also see the 512 bit memory bus finally get worked like it should. Oh yes, and at the high end we can finally expect to see 1 GB of GDDR-4. If AMD/ATI can actually implement these changes (easier said than done), then we could actually expect to see a product that would potentially outperform the high end 8800 Ultra. AMD/ATI would have a several month lead on NVIDIA's successor to the 8800, which is expected in November. Hopefully, they can get it done this time! EVGA 8800 Ultra Superclocked While browsing around this weekend, I noticed that EVGA has released a Superclocked version of their 8800 Ultra. While most were not particularly impressed with how well the stock clocked Ultra performed, it seems that EVGA took the criticism seriously and has released what looks to be a monster Ultra. It sports a 655 MHz core clock, memory at 1125 MHz (2250 effective), and shader clocks sitting at 1663 MHz. This is a significant jump from the regularly clocked Ultra! I have not seen this card online yet, but expect the price to initially hit above $850. XFX also features an overclocked Ultra, but it is not quite as fast. MSI on the other hand has an Ultra for $799 that is clocked at 660 MHz core and 1150 MHz memory, but I am not sure what the shader clocks are at, which is probably more important than the extra speed the core and memory have over the EVGA product. One other advantage that the MSI product has over the EVGA is the inclusion of the title Company of Heroes, which has gotten rave reviews (and I still haven't tried it yet). I still like bundles. It is good to finally see some work being done on the Ultra lineups, and it certainly seems like there are going to be plenty of these cards running around as compared to the GeForce 7800 GTX 512. May 25, 2007 Gigabyte GeForce 8600 GTS Silent Pipe 3 Review - Josh For the past few weeks I had the pleasure of testing and gaming with the 8600 GTS. While it was released to a lukewarm reception, I have come to discover a lot of hidden potential and surprising performance from this little card. Gigabyte goes the extra mile in providing a silent cooling solution that actually works quite well. Throw in a copy of Supreme Commander and a street price under $199 US, and you have a very tempting Direct X 10 upgrade. Here is a quote:
You can read the entire review here. Mike Hara and NV's Next Gen - Josh Over the past few years I have communicated quite a bit with Mike Hara. He is a Nebraskan native, and since I live in the bordering state of Wyoming, we often had quite a bit in common with our distinctly western outlook. This is a big change from most of the other people in the industry who typically range from the states of Texas and California. Mike is a pretty straight shooter, which is probably a good thing since he is VP of Investor Relations. As such, he can be surprisingly candid in some of the details he releases. One of my first experiences with this was when Mike told me that the upcoming GeForce 3 was going to be a 60 million transistor monster, more than twice the size of the previous GeForce 2. That information was not well known, nor was it officially verified until the launch of the GeForce 3 four months later. Well, Mike dropped another bomb recently in an investor meeting, stating that NVIDIA's next generation high end products will perform at the 1 teraflop level. If we consider that the current G80 based 8800 GTX runs at around 460 gigaflops, we can see that come November the 65 nm beast NV undoubtedly will come out with will be around twice as powerful. Now, that is in pure computing power, and not necessarily performance. We still have issues such as potential fillrate, memory bandwidth, and driver efficiency to deal with. Here is what I figure we can expect. I would be willing to bet that NV will integrate another 32 scalar units into the high end design. These stream processors will then run at 2 GHz. The overall die size will most likely be about the same as the current G80, if not a little smaller. NVIDIA would then also scale the texture and ROP units with the two new SIMD units. Core clockspeed would also see an increase to around 700 MHz. GDDR 4 would be much more available in that time frame, so NV would probably stick to the 384 bit bus for the high end, and the slightly cut down 320 bit bus for the lower-high end card (8800 GTS on steriods). We can also expect the usual "low level architectural improvements" which will positively impact per clock performance. If NV were to do such a thing, we could expect power consumption to be at the same level as it is now with the 8800 GTX, or even potentially a slight bit lower due to the advances in leakage reduction in TSMC's 65 nm high speed process. Mike gave an interesting hint of what to expect, and what the potential timeline is. Unless he is pulling a "David -we are not yet going to unified- Kirk" and sowing a bit of disinformation, AMD should be able to release their 65 nm high end refresh in late August/early September and have a bit of an advantage over team green. That is, if AMD can work out a lot of the kinks in the R6x0 architecture. In that case I would expect nearly the same amount of scalar units (5 x 64), but an increase in ROP and texture units. Hopefully AMD will also fix the current ROPS so that anti-aliasing performance is not so dismal as it is now. Work on the scheduler would be a huge improvement, as it seems the 320 scalar units are not being fed as optimally as they could be, especially in current applications. If AMD can improve these things, plus scale clockspeed to the 850 MHz to 900 MHz mark, utilize the 65 nm lower leakage advantages, then they could have a pretty mean and much more power efficient product on their hands as compared to the HD 2900 XT. Oh yes, let's not forget about including UVD this time into their high end product. May 22, 2007 Auzentech X-Meridian Review - Josh Around the first of this year Auzentech offered to send me their high end soundcard for review. Being a part-time audio enthusiast, I readily agreed. This has turned out to be one of the cleanest cards I have ever listened to, and if I were in the market for a soundcard, this would be one I would take a keen interest in! While C-Media still struggles with the impression of being a second rate soundchip vendor, the X-Meridian should turn a few heads and perhaps cast C-Media and their CMI based chips in a different light.
You can read the entire review here. May 20, 2007 GeForce 8600 GTS Impressions - Josh I finished testing today on the 8600 GTS, and I must say that it is an interesting bird. On release day many sites panned the card saying it didn't offer enough performance for its price range, especially considering some of the competition (7900 GS and X1950 Pro). Well, we have a newer driver, and I had a chance to test it against the old 7800 GTX (essentially the same performance as a 7900 GT). The results were somewhat surprising, and definitely enlightening. There is more to the 8600 GTS than meets the eye, and I have a multitude of benchmarks that prove it. What is perhaps more interesting about this review is that the card itself is the Silent Pipe 3 8600 GTS from Gigabyte. This design is fanless and it features a very complex cooling system. I have taken temps and watched power consumption quite closely. This is also one of the few 8600 GTS cards without a PCI-E connector to supply it with more juice. My detailed review and conclusions on both the technology behind the 8600 GTS, as well as the video card itself will be up this week. May 17, 2007 HD 2900 XT First Impressions - Josh Yesterday the fine people at FedEx allowed me to pick up a certain box from Canada at their local depot. Upon opening the box I was in for a small surprise. The HD 2900 XT sample I received was actually a retail model. I was expecting an anti-static bag with the card and a CrossFire connector, and perhaps a burned CD with drivers. Nope, it was the full meal deal. ATI packed the thing with plenty of components, but it was lacking one thing... no extra 6 pin or 8 pin power cable converters. If you buy a HD 2900 XT, make sure your power supply has at least 2 x 6 pin PCI-E connectors, or you will not get it to work. The retail package does include the HDMI adapter, which is pretty spiffy in my opinion. AMD already has HD 2900 XT supporting drivers on their website, and apparently these are very similar to the ones that AMD provided the press at launch. So far the few applications I have been able to run seem to do well, but more in depth testing is of course needed. The noise for the most part is not as bad as I was expecting, and when it first boots up the card is pretty much silent. Even when running in Vista with the full Aeroglass treatment, the card doesn't spin up the fan at all. The only time you start to hear the fan is when playing a 3D game or running a 3D benchmark. Even then the noise actually seems to be less annoying than the older X1900 XTX/XT cooler. When running a 3D app, the back of the card right behind the chip shows a temp of approximately 51C, which isn't nearly as bad as I had expected. With people reporting temps of upwards of 90C, one would expect the back to be very close to that number. I wonder if we are seeing a newer revision of the chip with the retail boards? It could very well be that the initial press samples were using an older revision of the chip/board. I will attempt to find out more about this later. The exhaust while running 3D apps is also not as hot as I expected. Not a whole lot more to say about the card yet, but so far I have been pleased by what AMD has delivered. It may not be a giant killer, but it certainly has the potential to be a nice product. It is packed with features, and with some more driver tweaking we can expect something that will likely sit right between the 8800 GTS and 8800 GTX in the majority of applications out there. May 16, 2007 AMD promised a MSRP of $399 for their HD 2900 XT, and bundled in that price is a voucher that allows users to download and play Half Life 2: Episode 2, Portal, and Team Fortress 2. Nobody is quite sure the retail value of "Black Box", but it certainly would be in the $29 to $39 range. Still, most retailers right now are offering their cards at around $411 to $429, just a bit higher than the AMD recommended $399 MSRP. Obviously retailers are extracting the "early adopter tax" from the people who are willing to pay it. Still, an extra $30 to the pricetag is not exactly pocket change, but it could be a lot worse. On NVIDIA's side, things are pretty interesting. I believe that NVIDIA marketing are having a heyday with the HD 2900 XT launch. Most everyone is talking about "power, heat, and noise" as well as accusations of "Quincunx 2.0" in regards to AMD's CFAA filters. Every reviewer has basically stated that the HD 2900 XT does not offer much more performance than the 7 month old 8800 GTS, which is available in mass quantities at price points from $259 for the 320 MB version and $349 for the 640 MB version. Now, it is not that these things were really pushed out in a PDF to reviewers (ok, maybe a few of them were), but for the most part these things are true. The HD 2900 XT, at this point of time, is really a competitor to the 8800 GTS. And because of its bundle, newness, and extended features (HDMI, UVD, etc.) it does cost more. I would guess within a week we will see the HD 2900 XT cards come out in greater numbers, and we will see the price start moving down to the $399 level and below. At the high end NVIDIA continues to sit pretty. The 8800 GTX has not gone down in price for several months now, and is sitting at a pretty reasonable $529. MSI released a highly overclocked 8800 Ultra board for $759, but as of today that price has gone up to $799 at most retailers. If you are looking for an Ultra, that would be my first choice, as the rest of them are still sitting at $829 and above. I think overall people now really appreciate what NVIDIA was able to do with the G80 chip. It is an impressive piece of technology, and while not perfect, it certainly is still the best out there. People who are early adopters of this technology must now feel that their investment back in Nov/Dec has certainly paid off, as prices have not gone down all that much comparatively, and the AMD HD 2900 XT has not knocked the 8800 GTX off its throne. I was notified yesterday that I will be getting a HD 2900 XT (or two) in from AMD, and I am really looking forward to testing it out myself. Seeing is certainly believing, and I do honestly think there are a lot of positives to the HD 2900 XT. If AMD had somehow been able to release this part back in November, it would have been a lot more warmly received. But because it is late to the party and really only competes well with the 8800 GTS, its reception has been much more muted (and negative in some cases). One thing that people are really starting to talk about is how it may perform in DX10. I remember way back in time when I was actively defending the GeForce FX series, I mistakenly thought that the FX series would really be able to spread its wings in DX9 applications with lots of complex code. I believed that the architecture of the FX series was more able to handle these complex instructions better than the R3x0 series of chips from ATI. Frankly, I was wrong. While the FX series was able to handle more complex instructions than the Radeon 9 series, it simply did not have the horsepower to do so effectively. I think we are starting to see a bit of this mentality leak into the HD 2900 XT supporters. Wait until DX10! Well, while workloads are different in DX10/WGF 2.0, I think that overall performance as compared to the G80 will reflect DX9 performance. So, while DX10 will bring some nice features and performance to the table when talking visuals, it will not miraculously enable the HD 2900 XT to suddenly compete with and overcome the GeForce 8800 GTX. While I believe we will see the overall performance of the HD 2900 XT improve over the next few months due to work done on the realtime compiler in the driver, again we should not expect miracles.
If you have found this article interesting or a great help, please donate to this site.
Copyright 1999-2007 PenStar Systems, LLC. |
|||
