Original Link: http://www.anandtech.com/show/2347
"Half our life is spent trying to find something to do with the time we have rushed through life trying to save."A plausible quote today from the great American Humorist, Will Rogers, but certainly not as famous as Clara Peller's quote in the main title while getting her 15 minutes of fame at Wendy's. It was on September 18th, 2007 that we started our X38 performance preview with this tag line: "The launch of the Intel X38 chipset is just a few days away and we have a feeling the product managers and press relations personnel in and around Santa Clara are a bit nervous at this point...." Were we ever wrong about that date, though the product managers and press relations personnel being nervous can't be far from the truth. As it turns out every previous date we have suggested, confirmed, or published concerning the X38 launch has been incorrect.
Why? There are a myriad of reasons but it all comes down to a lack of communication and planning in our opinion. It seems as if everyone was on the X38 Express bus, but sadly nobody was driving it correctly. Sure, our favorite drunk uncle was taking us for a joy ride in August, then our bipolar sibling had the wheel in September, and we entered October with our driver's education challenged child at the wheel.
Well, here we are on October 10th and the bus has finally pulled into the station. Unfortunately, it seems we have lost a few passengers on the way. We looked around the bus as it pulled in and noticed the usual website reviewers, a few engineers, a couple of marketing people with their heads still spinning, and Mr. Gigabyte and Mr. ASUS were still on board - though with a weary and beaten look. Apparently the ladies and gentlemen from MSI, Foxconn, ECS, DFI, abit, Biostar, and even Intel were missing, just like their X38 products are today. Did they jump ship for safety, or were they thrown from the bus? Maybe a little of both....
We pulled out our notes from Computex (early June 2007) and started looking at the dates, with the first one being from directly from Intel that stated the X38 and G35 were launching within 90 days. Fair enough, and we even heard rumors that it was really looking like the end of July based on how well the initial samples were performing. We tried to get an official date and were told late August to early September was still doable. Privately, during June, we still were being told late July. July came and the new schedules were late August or so, exactly what we were officially told at Computex.
Near the end of July we started to hear whispers of a delay - a minor delay at that - but still in the September time frame, though later in the month. This was completely understandable; everyone has delays (including yours truly being double guilty as we speak). Our first sample boards arrived in early August from several manufacturers and off to work we went on the shakedown phase.
The boards appeared to be very solid, but had some of the typical problems you see early on in the process. Our major concerns revolved around the lack of noticeable performance improvements in overclocking and memory performance over the quickly maturing P35. These are two points that Intel had made sure to emphasize as a selling point of X38 for the enthusiast market; however, everyone was convinced that the upcoming BIOS improvements would make a difference. It turns out BIOS updates would help to some degree, just not in the time frame we expected.
Finally in late August, all of the possible release dates we had been given for several weeks were officially nailed down to September 24th - well it was the 17th but that is another story. The manufacturers were comfortable with having supply on that date and some even envisioned the P35 launch situation where boards were in user hands before the launch and perhaps even before the review sites had them. We felt safe enough to provide a preview of the Gigabyte GA-X38-DQ6 on September 4th and all was still clear at that time for a 9/24 launch.
Little did we know that the silicon we were testing was about to be changed due to some technical issues found at the last minute. Not only was the X38 chipset being revised, but it appeared in most cases the board components would need to be changed as well. This left the manufacturers scrambling (and Intel as well). Initially the launch on the 24th still had a green light, but it would occur in a limited fashion, just like what we have today. We found out right after our ASUS preview article on the 18th that Intel decided to move the launch date to October 11th. We were disappointed, as several weeks of testing, feedback, and work at the expense of other activities left us wondering about the whole process. It was about to get a lot worse.
It seems that although the newly revised X38 and board layouts were basically ready by late September, the BIOS microcode was still undergoing significant changes and chipset production had not ramped back up quickly so supply was going to be tight for the Tier One suppliers and even worse for the Tier Two companies. Another week went by and confident that the release schedule could be improved, the performance NDA embargo was moved back to October 9th or 10th depending on your time zone.
However, this date was a shot in the dark as it turns out only ASUS and Gigabyte have product ready and it is slowly being introduced into the channel at this time. Besides supply being limited, so is choice - something we tend to take for granted in the vast world of PC components. Invariably, this leads to high prices, frustration, and sometimes being the guinea pig at new product launches.
A few who have bought the X38 boards early have probably experienced those feelings, as the initial BIOS spins are not up to speed in regards to performance or compatibility. Intel has been working diligently - at times frantically - to optimize the BIOS microcode and assist manufacturers with optimizing their X38 boards. We have seen this pattern play out several times, in particular the P965 launch is similar in this regard, but we thought Intel had nailed the launch process with the P35. Unfortunately, we in turn got nailed with the X38 launch process, as did many early adopters of the product.
Our preview today is just that: we are only discussing the chipset specifics and not showing board results, at least yet. The reason for this is that over the course of the last five days, we have received a total of seven BIOS releases for the two retail boards we have on hand. We tested five of these releases and let the manufacturers know our thoughts about them. Up until the last BIOS updates a day ago from Gigabyte and ASUS, our reviews of their flagship products were not going to be very positive. It is one thing to show progress in a preview article and talk about potential, it is a whole other subject when it comes to released product.
However, since the last BIOS releases arrived right before the chipset was officially announced, we decided to test and report with these BIOS spins that will be publicly available. We have already noticed several improvements ranging from 4x1GB compatibility being greatly improved to performance enhancements that are starting to show the strength of the X38 chipset. By no means are these BIOS releases perfect yet, but at least we have the opportunity now to show results with various components and memory configurations - several of which did not work correctly or at all just a few days ago. In fact, a few of the BIOS releases from last week seemed to make the boards worse, and we could tell they were rushed out the door to meet the launch schedule.
That brings us to the review candidates, which total three boards at this time. We have the Gigabyte GA-X38T-DQ6, Gigabyte GA-X38-DQ6, and the ASUS P5E3 Deluxe, and reviews will be published over the coming days. Additionally, the ASUS Maximus boards have shipped and should arrive shortly. Where are the other boards you might ask? Well, it appears from the chipset delays, BIOS microcode changes, and supply schedules that boards from the likes of abit, MSI, and Foxconn should show up in the next couple of weeks with others hitting the market in early November. In fact, Intel's own Bone Trail board now has a target date of "Later in Q4". And that just about sums up this whole launch process.
We have the familiar Intel block diagram that basically outlines the same technology patterns in previous MCH/ICH families. The X38 MCH is paired with the ICH9 series of Southbridges that was introduced with the P35 chipset. Intel continues to utilize its Direct Media Interface (DMI) technology for the interconnect link between the ICH/MCH chipsets. The 2GB/s DMI capability has not changed with this latest offering and continues to offer more than enough bandwidth for most users.
We say "most users" as installing a six drive RAID 10 array, audio card, Turbo Memory card, and TV Tuner card along with fully utilizing the LAN and USB ports can provide enough interconnect clogging data that even the most hardened air traffic controller (i.e. chipset) would fly into a panic. However, that particular scenario is very rare on the desktop. Intel will be abandoning the DMI interface late next year when the second generation 45nm CPU, Nehalem, is introduced. Intel's CSI (Common System Interconnect) and IMC (Integrated Memory Controller) represent their long-awaited response to AMD's HyperTransport technology and do away with the current front side bus architecture.
The most significant changes in the X38 MCH over the P35 or 975X are the inclusion of two PCI Express x16 lanes, PCI Express 2.0, and official support for DDR3-1333. One of the features in the X38 MCH that Intel has been fairly quiet about is the revised Snoop feature. It's not what you think, as we can confirm that Snoop Dogg was not involved in this development nor is it a feature that lets NFL head coaches steal signals, or the NSA figure out what you had for breakfast.
While Intel's press slides simply tout the virtues of Intel's "faster memory access", the fact is Intel has revamped the X38's memory controller to include their revised "flexible clock crossing architecture" and improved prefetching circuitry. We have done a little digging into exactly what faster memory access means and it seems to revolve around the snoop cache buffer improvements. The snoop feature can be described very simply as an extra level of pseudo-cache resident in the MCH. It's not really cache, per-se, but lines of recently cached memory reads. The purpose of this feature is to reduce memory read latencies in memory intensive programs that read a lot of data in parallel. The most common scenario is in multiple core computations where the cores are all manipulating shared data a little differently.
The primary benefit is that a separate memory read isn't needed every time data is accessed; instead the X38 MCH intelligently caches data and provides it when available. This also allows for a larger re-order buffer for read and write operations that further enhances memory performance. Every time the MCH must switch between reading and writing there is a wait period on the data lines, so the X38 MCH will wait until there are enough write requests stored before committing the data to memory.
Essentially, the MCH will "store" write requests and then "burst" them into memory when either the filter is full or the core requests data from memory that has not been written yet. So when the MCH writes this data it is coming straight from the pseudo-cache and this is what provides the "faster memory access" in the X38. It is almost like the chipset engineers followed their CPU brethren in the op re-ordering design in the Core 2 series, only it is applied to memory read/writes and implemented by the MCH - or maybe it's just a highly refined feature from the Intel 870. In reality, we have noticed slightly improved latencies and read rates when compared to the P35, but nothing that would make us dance in streets.
Along with support for both DDR2 and DDR3, the X38 introduces official support for DDR3-1333 while the P35 only officially supports DDR3-1066 - even though we have not had any issues running P35 boards past DDR3-2000 with the right memory. Intel is also introducing their Extreme Memory Profile (XMP) technology with the X38 roll-out. XMP is just like the Enhanced Performance Profile (EPP) technology launched by NVIDIA last year. It simply is a means of adding additional memory timing and clock speed profiles to the DDR3 SPDs in the same way EPP does for DDR2 memory. These profiles are designed to make it easier for users to basically auto-tune or overclock their memory/system using specific XMP profiles instead of manually changing individual timings and bus speeds in the BIOS.
Fortunately for those wanting to upgrade to the latest Intel chipset and not needing a new home equity loan to purchase DDR3 modules, the X38 MCH also supports DDR2 memory. In early testing with retail boards and BIOS releases, we are not seeing any real improvements over the DDR2 based P35 boards. Since the X38 is not "optimized" for DDR2 memory operations, this is both good and bad news. However, until DDR3 memory prices subside, an upgrade from the P965 or 975X should not be too painful once boards are in plentiful supply.
The last new feature from Intel is the support for the PCI Express 2.0 standard. In fact, Intel has the first desktop chipset on the market that supports this new standard. (AMD's RD790 should appear in the not-too-distant future.) The big news is the PCI Express 2.0 specification doubles the interconnect bit rate from 2.5 GT/s to 5 GT/s per port and is completely cross-compatible with the 1.0/1.1 specifications.
This means several things. First is that the performance increase to 5 GT/s effectively increases the aggregate bandwidth of a 16-lane link to approximately 16GB/s (maximum theoretical bandwidth of 8GB/s in each direction simultaneously before overhead), double that of the 1.1 spec. Real bandwidth per lane will be up to 4Gb/s (estimated to be around 500MB/s per lane/pin on average in current testing) in each direction given the 8b/10b encoding method used to transmit the data.
This increase in bandwidth comes courtesy of faster signaling rather than wider data paths which is why a 2.0 card is compatible in a 1.01/1.1 slot and vice versa. However, a 1.0/1.1 card will only work at its rated speed in a 2.0 slot and a 2.0 card is limited to the 1.0/1.1 slot speed. When two 16-lane PCI Express ports are utilized, the second port will support PCIe 1.1 cards at x8, x4, or x1 speeds or PCI Express Graphic cards at x16 or x1 operation.
While most will concur the reasoning behind the upgrade to the PCI Express standard is to improve graphics bandwidth, we think several other revised features also played a part in the early adoption of this specification. These features include dynamic link speed management, link bandwidth notification, access control services, and the power limit redefinition protocol. Of these, the dynamic link speed management and power limit redefinition are the two most interesting features in our opinion.
The dynamic link speed feature includes support for software controls that can dynamically throttle lane speeds. The power limit redefinition feature allows the system to redefine the slot power limits based upon the device inserted into that slot. The latter feature will work well with the new 300W electro-mechanical (CEM) spec. This new specification that works on either PCI Express standard provides full support for the 8-pin auxiliary power connectors seen on video cards like the HD 2900 XT and upcoming NVIDIA G9x offerings. The 8-pin PCIe power connector is capable of delivering up to 150W of power compared to the 75W limit in the 6-pin PCIe power plug. The PCI Express x16 slot on the motherboard is still limited to 75W. In total, up to 300W is available for each x16 PCIe slot on the motherboard, and hopefully we will not reach the day where that capability will need to be increased. (The HD 2900XT design had us wondering for awhile....)
The current range of performance chipsets on the market include NVIDIA's 680i SLI, Intel's P35, and the venerable 975X that is finally being retired after a two year run. Quickly glancing over the specifications we see that Intel is the only provider of DDR3 memory support, though NVIDIA is working on a DDR3 compatible version of their upcoming 780i chipset. The X38 Express MCH offers a total of 32 PCI Express 2.0 lanes, which can be split into two PCIe x16 slots for multi-GPU configurations such as CrossFire. Although the chipset is fully capable of supporting NVIDIA's SLI technology (it runs it very nicely we might add), there will not be support for it within NVIDIA's drivers at this time.
The 975X offers a total of 16 PCI Express 1.0a lanes that can be split into two x8 slots while the P35 only officially supports a single x16 slot with its 16 PCI Express lanes. NVIDIA relies on its 680i SPP and MCP to each provide the necessary amount of PCI Express lanes for its dual x16 capabilities. As such, it requires significantly more interconnect bandwidth as graphics data is traveling across its HyperTransport based interconnect link.
The official FSB speeds for the X38, P35, and 680i SLI are listed as 800/1066/1333 with the 975X being limited to 800/1066 - although with the right board we have not had problems with the 1333FSB processors. However, the same cannot be said of the 680i SLI boards; while the chipset officially supports 1333FSB rates, we have seen several early board designs that lack the required hardware to fully support the 1333 capable CPUs, especially the QX6850 quad-cores. The same could also be said of the first generation 975X motherboards as they required significant component changes to fully support the Core 2 Duo processor series.
The X38 Express MCH also has the highest current TDP ratings, stating 26.5W when utilizing DDR3 memory and slightly more with DDR2 memory. The original numbers given to us by the manufacturers had pegged the TDP rating around 36W, but Intel's official response this afternoon has dropped it to 26.5W. NVIDIA has not published their official TDP ratings, though we estimate it be in the 16W~18W category for the 680i SLI SPP.
G35 Chipset Overview
We should see the introduction of Intel's G35 chipset in the next 30 days, which will be Intel's most advanced integrated graphics chipset and replaces the G965.
The G35 will be based on Intel's next-generation X3500 Media Graphics Accelerator. DX10 support is claimed, including full support for Shader Model 4.0 operations, although Intel does not expect to have driver support ready until late Q1 2008. (That might be a familiar tune for anyone running G965 X3100 and wondering about full SM3.0 support for Vista.) For backwards DirectX 9.0c compatibility the X3500 will also support Shader Model 3.0 and Hardware Transform and Lighting (T&L). X3500 will also offer native HDMI 1.2 and DVI output with HDCP support for HD-DVD and Blu-Ray playback at up to 1080P. MPEG2 and VC1 hardware acceleration is supported by the Intel Clear Video feature of X3500. Like the other series 3 chipsets released this year, all current Intel processors are supported as well as future 1333FSB Penryn 45nm processors.
We also have some information about the mysterious X48 chipset, which may help to quell rumors. First, we do not expect to see this chipset until the end of Q1 2008 based on current roadmaps. Yes, the initial rumors about the chipset showing up in November were based on other roadmaps (the details are always subject to change). The plan was to offer it in the high-end enthusiast sector, catering to those who overclock for a living or can afford the most expensive of components.
However, with the push back in the X38 schedule and the fact that several boards might not even be available until November, this "ghost" of a chipset has been delayed. What is the X48? It's basically an X38+ to be honest. It will officially support 1600FSB (which X38 unofficially supports), offer improved memory and overclocking performance, and in some corners of the universe it might even bring world peace. We understand it to be the Racer X version of Speed Racer and nothing more.
The X38 Express chipset is "officially" launched as of today and it brings a great deal of promise with it for the enthusiast user. However, as impressed as we are with the specifications and some early performance results with the latest BIOS releases, this launch has been anything but smooth. Intel has been executing extremely well this past year in all phases of their business, but the launch of their premiere performance chipset has been totally botched in many areas.
It has been a very disappointing process for us and even more so for the manufacturers from what we understand. We can only wonder what the early adopters of the boards think, but we will say that problems are being fixed quickly now. Hopefully the pitfalls in this launch process do not foretell what we will experience with the upcoming Penryn roll out. If it does, AMD might get some positive press at the end of this year when 790FX and Phenom are introduced.