A Lesson in User Failure: Investigating the Serial ATA Connector
Something you learn quickly in this industry is that working as a technology journalist does not make you immune to computer problems or the laws of physics that can be at the root of them. Just by doing our work we tend to break things now and then; overclocked processors become keychains, overheated video cards become surgical knives (make no mistake, PCB is a very capable blade), and gadgets become interesting conversational pieces. Much of this we'll make a passing observation on, but otherwise we don't talk about failures too often.
Every once in a while though, we will break something in a process that's genuinely interesting. Failure is its own reward, it teaches us how to not do something or do something better than we did before. And in those handful of cases, we like to get to the bottom of what went wrong, what we did wrong, and what can be done to avoid the issue in the future. In these cases, you the reader can receive some of our imparted knowledge without needing to also experience the pain and cost of the lesson.
So what have we managed to break this time that we find so interesting that it's worth writing about? We made what is in fact a very common mistake, and nearly turned a week-old hard drive in to a new source of magnets by breaking the Serial ATA connector on the drive. It's the kind of problem that sounds rather trivial, but due to the construction of many SATA hard drives, breaking the SATA connector is a death sentence for the drive because it's impractical-to-impossible to replace it, as it's part of the circuit board if not also part of the drive itself.
It's only appropriate to preface this by saying that we're not dissatisfied with the SATA specification, rather we find ourselves in an interesting situation. The thinner cable is far easier to route in a cramped case than a Parallel ATA cable, it doesn't impede airflow like a ribbon cable, and getting rid of hooking two devices to a single cable was a long-overdue change.
But - and we know we're not alone in this thought - SATA cables and connectors aren't quite as robust as the old PATA design. PATA cables could be worked in to rather impossible situations as the connector was extremely snug fitting, and the cable itself was extremely flexible when it needed to be folded longitudinally; it was hard to set up but also hard to break. We'll still take a SATA setup any day of the week, but we've come to the realization we can't abuse SATA setups like we could PATA setups.
As a consequence, today we'll share with you what we found out in dealing with our problem. What did we do wrong? What can we do about it? And just why is the SATA connector designed the way it is anyhow? Read on to find out.
Saturday, January 19, 2008
In Memory Of The Law: The Memory Industry's Legal Problems
In 1890 the United States government passed the Sherman Antitrust Act, a law that formed the bedrock of the United States' policy against monopolies and other unfair forms of competition. In the many years since then, further acts have been passed to amend the law, and the Government has in time made the correction of anti-competitive actions one of its more important roles. What started with the breaking up of truly gigantic corporations like Standard Oil and AT&T has moved on to include dealing with cartels that while not of a single corporation, act at times in manners similar that result in anti-competitive actions taking place.
If we were to take a cursory glance at the computing industry as a whole and try to pinpoint the areas where anti-competitive legal issues were likely to occur, we'd look at areas like CPUs and GPUs, where only a couple of serious competitors exist in each, or operating systems, where Microsoft has and continues to have a de facto monopoly. While these areas do in fact have issues, we would be missing an area that has shown some of the worst behavior. It turns out that the memory industry is one of the greatest offenders.
Generally speaking, it's counter-intuitive to see the memory industry as being a hotbed of legal woes due to the highly competitive nature of the market. With the JEDEC association setting very rigorous standards for RAM, products are quite literally perfectly competitive (from a non-overclocker's point of view): a part specified to meet a certain JEDEC RAM standard should be just as good as any other part that adheres to the same specification. As a result, OEMs can and do switch RAM on a regular basis depending on who can supply it at the lowest cost, and as a result we usually see the memory market operate as it is: a highly competitive market in which multiple companies supply the same good.
But didn't we just say that the memory market is one of the greatest antitrust offenders? Yes, for in spite of the memory market being very cutthroat, it's also an industry that is subject to highly volatile demand. It can be extremely profitable as a result when demand is far outstripping supply and it takes years to bring new fabs online to produce additional memory. It's the profitability of these periods that keeps nearly a dozen major companies in the business. With such volatility however, it also opens the window for manipulation of this volatility - if now is not a boom year, why not make it one?
Over just the last decade, the memory industry has been through no less than three major shakeups. The first is the infamous and now settled Rambus patent case, started in 2000 when Rambus asserted that it held patents on technology used in DDR RAM and wanted royalties as such, only to be found guilty of breaking antitrust and deception laws in acquiring these patents and covertly trying to influence the memory market. The second case involves the "Big 10" memory manufacturers colluding to keep RAM prices artificially high between 1998 and 2002, to which they were found guilty. Finally, a new investigation has opened up as of this year into the flash memory market, where the Justice Department is trying to figure out if there is evidence of collusion and price-fixing there too.
Today we'll be taking a look at the latter two actions, one just wrapping up while another begins. What exactly went on in these cases? How are or potentially were consumers hurt by all of this? What has been done to punish the offenders and to correct the market? Let's find out.
If we were to take a cursory glance at the computing industry as a whole and try to pinpoint the areas where anti-competitive legal issues were likely to occur, we'd look at areas like CPUs and GPUs, where only a couple of serious competitors exist in each, or operating systems, where Microsoft has and continues to have a de facto monopoly. While these areas do in fact have issues, we would be missing an area that has shown some of the worst behavior. It turns out that the memory industry is one of the greatest offenders.
Generally speaking, it's counter-intuitive to see the memory industry as being a hotbed of legal woes due to the highly competitive nature of the market. With the JEDEC association setting very rigorous standards for RAM, products are quite literally perfectly competitive (from a non-overclocker's point of view): a part specified to meet a certain JEDEC RAM standard should be just as good as any other part that adheres to the same specification. As a result, OEMs can and do switch RAM on a regular basis depending on who can supply it at the lowest cost, and as a result we usually see the memory market operate as it is: a highly competitive market in which multiple companies supply the same good.
But didn't we just say that the memory market is one of the greatest antitrust offenders? Yes, for in spite of the memory market being very cutthroat, it's also an industry that is subject to highly volatile demand. It can be extremely profitable as a result when demand is far outstripping supply and it takes years to bring new fabs online to produce additional memory. It's the profitability of these periods that keeps nearly a dozen major companies in the business. With such volatility however, it also opens the window for manipulation of this volatility - if now is not a boom year, why not make it one?
Over just the last decade, the memory industry has been through no less than three major shakeups. The first is the infamous and now settled Rambus patent case, started in 2000 when Rambus asserted that it held patents on technology used in DDR RAM and wanted royalties as such, only to be found guilty of breaking antitrust and deception laws in acquiring these patents and covertly trying to influence the memory market. The second case involves the "Big 10" memory manufacturers colluding to keep RAM prices artificially high between 1998 and 2002, to which they were found guilty. Finally, a new investigation has opened up as of this year into the flash memory market, where the Justice Department is trying to figure out if there is evidence of collusion and price-fixing there too.
Today we'll be taking a look at the latter two actions, one just wrapping up while another begins. What exactly went on in these cases? How are or potentially were consumers hurt by all of this? What has been done to punish the offenders and to correct the market? Let's find out.
The Gigabyte GA-X48T-DQ6 - Redefining the High End?
Gigabyte's motherboards are always products we look forward to, especially the budget/midrange solutions. These boards represent features and performance that often belies their price. Recently, Gigabyte has also been making strides into the high-end of the motherboard market. The first product of Gigabyte's aspirations using an Intel chipset was the GA-X38-DQ6 motherboard. The board was generally well received and had a decent set of features and performance. In short, we would not hesitate to recommend the GA-X38-DQ6 for a top-end overclocking or gaming system.
The only gripe about the GA-X38-DQ6 by hardcore overclockers is that it has slightly lower clock for clock performance when compared to other X38 motherboards. The GA-X48T-DQ6 we are reviewing here today obviously uses Intel's higher speed binned X48 chipset. Instead of just plugging the X48 into the existing X38 board design, Gigabyte listened to the concerns of users. Gigabyte made a wise decision to refine the existing X38 motherboard design to take advantage of the additional performance potential of the X48 speed bin.
A few months ago our feelings about the X38 chipset were a little mixed; we felt it really did not bring anything exciting to the table. As time has progressed since the chipset launch, motherboards based on the X38 chipset have become our favorites for overclocking, for many different reasons. In fact, the primary reason is that they have proven to be extremely reliable for overclocking use on a 24/7 basis. Once set up correctly, we find these boards to be able to run the same settings day-in day-out, so long as the demands are reasonable. Naturally for the budget minded users, we would still lean towards the P35 chipset as far as single graphics card use goes. However, as the focus shifts, the prices of X38 based boards should come in lower than the higher end P35 boards, making the decision of which one to choose all too easy.
It really is no surprise that Intel CPU's are at their very best when teamed with Intel chipsets. Understandably, the release schedule of all the tier-one suppliers includes motherboards in either DDR2 or DDR3 format using either the X38 or the updated (speed binned) X48. While we have always felt that the synthetic performance figures of the X38 in DDR2 form have been lower than expected, the 3D performance gains over more attractively priced P35 chipset is always apparent. In DDR3 format the X38/X48 is the performance choice, and outperforms the DDR2 boards overall in just about every benchmark… well, at least by a few percent. Of course, this slight increase in performance comes at an expensive cost, with DDR3 memory prices being double that of DDR2 - if not more - depending on what speed bin you order.
A couple of weeks ago, we were able to provide a small glimpse of the high-end extreme benchmarking that the DDR3 based GA-X48T-DQ6 is capable of in current form. We managed a clean sweep of current single card 8800 GTS 640MB 3DMark world records using this motherboard. The board's overclocking performance impressed us, but using a motherboard in a normal operating environment like most users is always an important part of our testing. Things were still quite rough around the edges at the time of the preview. In fact, we were unable to install Microsoft Vista due to our boot drive being unrecognized as a valid partition after Vista had completed formatting the drive. Issues like these are not new to us; most of the boards always need BIOS updates in their early stages. Our real concerns at the time revolved around non-working memory dividers and general unpredictability when overclocking.
Things have certainly progressed in some areas since our first look; we have received a few BIOS spins addressing improvements and compatibility in several areas. In truth, this has not been the most solid pre-release board we have received of late. We were actually beginning to feel a little spoiled, as most of the X38 based pre-release boards we received have been remarkably ready for good overclocking right from the get-go. In spite of this, we decided to keep the length of time between the first look and our review as short as possible. This is especially important in light of the fact that we are endeavoring to provide users with meaningful BIOS insights before Gigabyte releases the board. Naturally, this process takes time, but we think it's well worth the wait considering the options available in the BIOS. Indeed, we are intrigued to see what Gigabyte has managed in bringing performance and reliability together in one package using the X48 chipset.
The only gripe about the GA-X38-DQ6 by hardcore overclockers is that it has slightly lower clock for clock performance when compared to other X38 motherboards. The GA-X48T-DQ6 we are reviewing here today obviously uses Intel's higher speed binned X48 chipset. Instead of just plugging the X48 into the existing X38 board design, Gigabyte listened to the concerns of users. Gigabyte made a wise decision to refine the existing X38 motherboard design to take advantage of the additional performance potential of the X48 speed bin.
A few months ago our feelings about the X38 chipset were a little mixed; we felt it really did not bring anything exciting to the table. As time has progressed since the chipset launch, motherboards based on the X38 chipset have become our favorites for overclocking, for many different reasons. In fact, the primary reason is that they have proven to be extremely reliable for overclocking use on a 24/7 basis. Once set up correctly, we find these boards to be able to run the same settings day-in day-out, so long as the demands are reasonable. Naturally for the budget minded users, we would still lean towards the P35 chipset as far as single graphics card use goes. However, as the focus shifts, the prices of X38 based boards should come in lower than the higher end P35 boards, making the decision of which one to choose all too easy.
It really is no surprise that Intel CPU's are at their very best when teamed with Intel chipsets. Understandably, the release schedule of all the tier-one suppliers includes motherboards in either DDR2 or DDR3 format using either the X38 or the updated (speed binned) X48. While we have always felt that the synthetic performance figures of the X38 in DDR2 form have been lower than expected, the 3D performance gains over more attractively priced P35 chipset is always apparent. In DDR3 format the X38/X48 is the performance choice, and outperforms the DDR2 boards overall in just about every benchmark… well, at least by a few percent. Of course, this slight increase in performance comes at an expensive cost, with DDR3 memory prices being double that of DDR2 - if not more - depending on what speed bin you order.
A couple of weeks ago, we were able to provide a small glimpse of the high-end extreme benchmarking that the DDR3 based GA-X48T-DQ6 is capable of in current form. We managed a clean sweep of current single card 8800 GTS 640MB 3DMark world records using this motherboard. The board's overclocking performance impressed us, but using a motherboard in a normal operating environment like most users is always an important part of our testing. Things were still quite rough around the edges at the time of the preview. In fact, we were unable to install Microsoft Vista due to our boot drive being unrecognized as a valid partition after Vista had completed formatting the drive. Issues like these are not new to us; most of the boards always need BIOS updates in their early stages. Our real concerns at the time revolved around non-working memory dividers and general unpredictability when overclocking.
Things have certainly progressed in some areas since our first look; we have received a few BIOS spins addressing improvements and compatibility in several areas. In truth, this has not been the most solid pre-release board we have received of late. We were actually beginning to feel a little spoiled, as most of the X38 based pre-release boards we received have been remarkably ready for good overclocking right from the get-go. In spite of this, we decided to keep the length of time between the first look and our review as short as possible. This is especially important in light of the fact that we are endeavoring to provide users with meaningful BIOS insights before Gigabyte releases the board. Naturally, this process takes time, but we think it's well worth the wait considering the options available in the BIOS. Indeed, we are intrigued to see what Gigabyte has managed in bringing performance and reliability together in one package using the X48 chipset.
Subscribe to:
Posts (Atom)