D-Link High Speed Home Network

Making Networking Experts of Us All

by Josh Walrath

September 29, 2004

3D Mark 2005 and Other News of Interest - Josh

For all of those waiting for the countdown to finish up, it certainly did so this morning at 8 am MST.  Futuremark Corp. proudly announces that it has released its latest benchmark for the graphics industry.  3D Mark 2005 is a bold step forward, as it does not support video cards before the PS/VS 2.0 generation of products (that means the GeForce 4 and Radeon 9200 series and below will not be able to run this at all).  3D Mark 2005 supports multiple rendering schemes, such as PS 2.0/2.0a/2.0b/3.0, as well as VS 2.0/2.0a/3.0.  This benchmark is of course quite strenous, even for the latest generation of high end cards.  Using the 65.76 drivers and the GeForce 6800 GT, I was barely able to get into the teens when it comes to FPS in this program.  The visuals are quite impressive, and with both ATI and NVIDIA having comparable parts in terms of technology and speed, I don't think we will see the controversy that 3D Mark 2003 brought upon the scene with the Radeon 9700/9800 and GeForce FX 5800/5900 generation of products.  I don't think the review sites can handle another round of driver optimizations and cheats!  The Futuremark site is screaming for mercy at this moment, but you can download a free copy of this benchmark from other fine mirrors like Fileshack.

NVIDIA has already piped up that its SLI enabled GeForce 6800 Ultra currently holds the highest result for 3D Mark 2005 (a whopping 7229 marks- while my single GeForce 6800 GT received around 4000 marks).  This of course was using the latest beta 66.51 drivers (which are not yet available to the general public) that look to significantly improve overall performance for 3DM05.  This driver is certified by Futuremark though, so what else can I really say?  If you are busy standing in line for this download and want to know what to expect, there are a number of "3D Mark 2005 Reviews" out on the web.  Here is a quick list of them.

Tech Report

Accelenation

Hexus

For those not terribly interested in 3D Mark 2005, but instead are looking for some budget PC3500 memory, then 3DXtreme takes a look at a couple of DIMMS from a company called SuperTalent.  These DIMMS have timings that can be considered "pretty loose" by the majority of users, but for somebody who wants the extra clock speed and not have to pay a whole lot for it, these might be right up their alley (or someplace else).  Perhaps I should stop there.

Marco and the kids at HotHardware (sometimes Really HotHardware) have posted a comparison between the lower-midrange offerings from ATI and NVIDIA.  These fellows face off the Asus Radeon X600 XT and the MSI PCX 5750 for a battle royale in the PEG slot.  I wonder which one pulls ahead in 3D Mark 2005?  I guess we can't know, since this review was released before the 2005 NDA!!!  Then again, Marco was mentioned in the credits of 3D Mark 2005...  Perhaps one day I too will be mentioned there (gazes longingly into the bright, clear, blue sky).

Howard at Neoseeker takes a good, hard look at the Gainward GeForce 6800 GT 2400/ Ultra Golden Sample (not Golden Shower!) which sports a non-reference cooling design that promises more than GeForce 6800 Ultra clockspeeds.  The 6800 GT has turned into quite the popular card with enthusiasts, as they generally clock to Ultra speeds without a problem (and look nice too, not to mention perform like a mofo at Doom 3- and its eventual licensees).  Neoseeker takes the hammer to it and see what comes out the other side.  Definitely worth a read.

Speaking of Golden Showers (I mean Samples), my favorite site to visit, NV News, has posted a review on the Gainward PowerPack! Ultra/2100 GeForce 6800 Golden Sample 128 MB (yes, Gainward loves to make us type far more than we are used to- how many times does a guy have to flip back and forth just to make sure he gets the stinkin name correct on the product!).  They take a step back into retro-land by comparing this product to the infamous GeForce FX 5800 Ultra.  Quite an interesting read, I would highly recommend it.  This product also features Gainward's non-standard cooling system, which performs quite well (but is a bit loud from what I am hearing... cough).

September 28. 2004

Interesting Items - Josh

Duane over at Gamer's Depot put up a good editorial talking about "where all the high end cards are".  It is a very well done editorial with official quotes from a variety of sources talking about where the GeForce 6800 Ultra's and Radeon X800 XT PE's are.  While the Ultra's are a bit more available, I have yet to really see a XT PE on the retail market.  The official story is that Dell gets every XT PE card made by ATI, and nobody else gets anything.  I have word from Abit that though they have a XT PE SKU, they have yet to actually receive product.  There is little doubt that these chips are hard to make (the NV40 and R420), as they are very complex and run pretty hot (as well as trying to get them up to the specified clockspeeds).  On the other hand, there are very good quantities of GeForce 6800 GT and Radeon X800 Pro's on the market, and any user should have a pretty easy time picking up either of those models.  Picking up a GT is a great idea, as they often overclock to Ultra levels (and beyond) without a problem.  Most of the X800 Pro's can run at XT PE clock speeds, but with one quad disabled it just can't quite make it in performance.  Some very interesting answers in Duane's article, so it is a good read.

3DXtreme takes a close look at the Albatron K8X800 Pro II version 2.0 motherboard, which is based on the VIA K8T800 chipset, and comes in the S754 flavor.  A very solid offering, with plenty of voltage modifications, but the lack of AGP/PCI locks tend to make overclocking a bit of a chore.  Overall a very positive review, but those looking for higher overclocks might want to pass on this board.

Marco from Club 3D wanted to let us know that they will be offering X800 XT PCIE and X700 Pro PCIE cards in the very near future.  Unlike many other manufacturers, Club 3D seems to have a very good relationship with nearly all 3D chip companies, and offer a wide selection of products for the European market.  It is almost unfortunate that they don't offer anything in the North American market...

My friend SH SOTN let me know that the latest leaked NVIDIA drivers have references to the NV48 chip.  This is quite a surprise for me, as I was not expecting NVIDIA to release anything more this year.  Very little is known about this refresh, but we can probably assume that it will be an optimized NV40 chip with native PCI-E support, as well as perhaps a solid clockspeed advantage over the current GeForce 6800 Ultra cards in the high end form.  It will be very interesting to see how NVIDIA places this part, as really only recently have they been able to supply their high end cards in good numbers to the market.  You would think that NVIDIA would want as high as possible a return on their NV40 products, since it was so stinkin expensive to develop in the first place.  From my understanding, ATI will not have a refresh part out this fall, as it appears the X700 series will be bearing the brunt of the marketing.  Plus, with the lack of high end parts from ATI at this time, it doesn't make much sense to introduce something even higher end than the current X800 XT PE.  Most likely the focus this fall will be high end PCI-E introductions using these currently available chips.

Finally, PCStats have released a review on the Gigabyte GA-8ANXP-D motherboard, based on the Intel i925X chipset.  As with most high end Gigabyte motherboards, this monstrosity is packed to the gills with features.  One look at the board and you will understand what I mean.  A total of 8 SATA ports are present, as well as all the other bells and whistles that Gigabyte tends to implement onto their products.  Definitely worth a read.

That is all the time I have for now, look for more reviews shortly!

September 27. 2004

AMD Athlon 64 3700+ Review - Josh

When AMD released its latest group of high end processors, they treated one of them as the "red headed stepchild" of the Athlon 64 family.  While the FX-53, Athlon 64 3800+, and Athlon 64 3500+ received the lion's share of publicity, a 4th product was released along with them.  The Athlon 64 3700+ was a high end processor aimed at the Socket 754 market, but due to AMD's focus changing to Socket 939, this processor was relatively unrepresented by the media.  This does not mean that this processor deserves this treatment, as it is nearly as fast as the FX-53 and 3800+ products it shares its clock speed with (2.4 GHz).  For any user with a Socket 754 based system, this processor could be the end all of performance for that platform.  Here is a quote:

            The Athlon 64 3700+ is based on the same core as the Athlon FX-53, except it does not feature the dual memory channel architecture.  It is clocked at 2.4 GHz with 128 KB of L1 cache (split into Data and Instruction cache) and 1 MB of L2 cache.  All 3700+ parts are from the CG revision, which has a slightly redesigned memory controller, HyperTransport controller, and other tweaks.  The memory controller does actually work at full processor speed, which can give it a significant advantage in some situations over off-die controllers (such as what the Athlon XP and Pentium 4 utilize).  Having the memory controller on-die also cuts the memory access latency down by a very significant margin.  While the 3700+ does not feature a dual channel memory controller, most memory accesses are not streaming accesses, and instead consist of one or two d-words.  This is the main reason why the Athlon 64 3700+ can keep up so well with the Socket 939 Athlons and the Pentium 4 processors.

You can read the entire article here.

September 24, 2004

VIA Releases AMD PCI-E Chipset - Josh

VIA is the first out of the gates with a PCI-E chipset, and there are actually samples in hand at many sites.  This shouldn't come as a surprise, as Abit has already shown its own K8T890 based motherboard sporting PCI-E functionality.  The first generation of motherboards based on this core logic will only feature the older VT8237 southbridge, which does not feature any PCI-E lanes or integrated Gigabit Ethernet.  There will also be a K8T890 Pro flavor, which will feature dual PEG slots.  This will not show up for a while though.  The eventual southbridge that will complete the functionality of the K8T890 will be the VT8251, which will feature an extra 2 PCI-E 1X slots as well as all the extra goodies.

Performance of this chipset appears good, but this is mainly due to the fact that the memory controller resides on the Athlon 64.  Initial tests were limited by what kind of video cards actually fit onto the motherboard (which essentially relegated all tests to the X600 series of cards).  VIA plans to offer these chipsets as a no-nonsense alternative compared to other manufacturers (read NVIDIA here).  These chipsets will be fairly cheap as compared to the NVIDIA products.  NVIDIA on the other hand will continue to add more features and functionality to their chipsets, but these will of course be a bit more expensive (but the integrated Gigabit Ethernet and hardware firewall are certainly compelling features).

VIA will also release integrated graphics versions of these chipsets, but they will have a new twist to them.  The Deltachrome architecture will finally be ported over to the integrated space, and this could very well provide a suitable alternative to the Intel 900 series of integrated chipsets.  If VIA and S3 can work out the kinks that have plagued the architecture and release a solid driver for it, then this could really help propel the Athlon 64 into the OEM market.

Another big question that should be posed is if PCI-E will make it to the Socket 754 products.  I am not sure what manufacturers will do here, but it seems somewhat unlikely for this reason: bandwidth.  PEG consumes 4 GB/sec of bandwidth each way, which gives a total of 8 GB/sec bidirectional.  The Athlon 64 939 processors feature the 1 GHz HyperTransport connection which translates into 8 GB/sec bandwidth, and the dual channel memory of the 939 processors equals 6.4 GB/sec.  As you can see with these numbers, everything is fairly well balanced.  When we turn to the Socket 754 chips, things become a bit different.  First off PEG still requires 8 GB of cumulative bandwidth, but the HT link now runs at 800 MHz giving 6.4 GB/sec of bandwidth.  While that bandwidth difference is not extreme, the main memory bandwidth is.  The single channel Athlon 64 runs at 3.2 GB/sec, which is far, far below that needed by PEG when it is running at full speed.  I may be making a mountain out of a mole hill here, and most likely a PCI-E motherboard for Socket 754 will hit the market (if past products from the Taiwanese motherboard makers are any indication).  Still, due to the bandwidth constraints, it might not be a very effective product and users may be well warranted to stay away from a Socket 754 PCI-E solution.  This is so far only my opinion, as I have had little chance to test out my theory.

Something else interesting to consider is that the Socket A architecture would be more able to handle PCI-E than the Socket 754 products, mostly because the main memory controller is located on the same chip as the PCI-E controller, and both VIA and NVIDIA have dual channel Athlon XP products out there already supporting 6.4 GB/sec of bandwidth to main memory.  This will most likely not happen as the Athlon XP is no longer "sexy" to the majority of users buying new computers.  Unfortunate, but true.

Here are a couple of previews so far:

Tech Report

PC Perspective

Hot Hardware

Other Interesting Things - Josh

Vivian from DFI let me know that they released a very interesting product for the Intel market.  One of the first hybrid i875/LGA 775 comes in the form of the LANParty 875P-T.  This will allow the use of a LGA 775 Pentium 4 processor with the mature (and really fast) i875 chipset.  For those not wishing to explore the DDR-2/PCI-E waters, this is a quick and easy way to transition to the new LGA 775 processors from Intel.  AGP video cards and DDR-1 DIMMS can be transferred over to create a fast and affordable system based on Intel products.  Definitely a product with a niche in today's market.

Crucial is giving away the Ballistix Dream Machine, which is an absolutely loaded customized machine.  Water cooling, DDR-2, 1 TB of disk space, X800-XT PE video, and a host of other goodies make this machine look amazing.  More details (and the sign up) can be found here.  You would be silly not to submit an entry for this machine.  On a side note, that thing must weigh a ton, especially when filled with water.

September 23, 2004

Thoughts on ATI's X700 Series - Josh

As you know, the embargo on ATI's X700 cards was lifted on Tuesday and we all had the chance to see what ATI's next generation of midrange products look like, as well as how they perform.  There will be three flavors of the X700, which allows ATI to address a significant portion of the market.  The top end will be the X700 XT, which will feature the core running at 475 MHz and 128 MB of GDDR running at 525 MHz.  The X700 Pro will come in 128 MB and 256 MB flavors, which run the core at 420 MHz and the memory (GDDR3) at 432 MHz.  The final iteration will be the regular X700 which features a 400 MHz core and 128 MB of DDR running at 350 MHz.  Prices will range from $249 to $149, depending on the model and memory loadout.

The primary review we saw was on the X700 XT model, which compares quite favorably with NVIDIA's GeForce 6600 GT.  The X700 series of cards are based on the RV410 chip, which is a derivative of the high end R420 (which powers the X800 series of cards).  As such it supports PS 2.0b, VS 2.0, 3Dc, and the integrated VideoShader HD.  The RV410 is made on TSMC's 110 nm process, which is a shrunk version of their 130 nm process.  This process does not feature Low-K, but it is quite impressive to see that TSMC's base process can support 120 million transistor products running at 475 MHz (and above in NVIDIA's case).  The new midrange has now caught up with the old high end in terms of performance, and we will start to see the old guard (namely the ATI 9800 series) be put out to pasture once X700 production picks up.

In overall performance the X700 XT tends to slightly lag behind NVIDIA's GeForce 6600 GT.  The differences are fairly minimal though.  The one area where the X700 has a leg up is in Vertex Shading performance.  While NVIDIA essentially cut the NV40 in half to create the NV43 (8 pixel pipes, 3 vertex shaders), ATI took a slightly different route.  While they cut the pixel pipelines down from 16 to 8, they kept all 6 vertex shaders intact.  In applications which require a lot of VS power, the X700 series should have a healthy advantage over the competition.  In most applications though, it seems that the 6600 GT is just slightly faster.  In terms of die size, they appear nearly equal.  The SM 3.0 functionality of the NV43 takes up a significant amount of transistors as compared to SM 2.0 functionality, but because there are 1/2 the VS units on the 6600GT it seems to balance out the two designs.  So, depending on the design and their yields, both ATI and NVIDIA seem to be paying about the same price for each of their respective products.

ATI has said that availability of the X700 and X700 Pro should be good within a few weeks, and the X700 XT will be some 4 weeks behind that.  Eventually AGP versions of this card will be available using ATI's bridge chip.  On the other hand, NVIDIA just announced that several retailers now have 6600 GT stock and are selling to the general public.  While supplies are somewhat limited now, within 4 weeks we should see very good availability (about the same time when we will see limited availability for the X700 XT).  If this truly is what is happening, NVIDIA is in a significantly better position going into the Holiday season than ATI is.  If ATI fails to deliver good quantities of X700 productby the end of October, then we will probably see NVIDIA take a significant chunk of the midrange market with their 6600 series.  What is more significant is that NVIDIA appears to be a lot closer to supplying AGP based 6600's to the market, which will probably become a popular upgrade product for people with older machines.  PCI-E, while gaining ground in OEM circles, is still fairly rare in the marketplace.  It is still unknown if either manufacturer will integrate the bridge chip onto the chip substrate, or just onto the PCB of the card.  I guess we will find out soon enough!

September 20, 2004

AMD Releases Thin and Light Athlon 64 and Other News - Josh

In what shouldn't be a surprise to any of you reading this site, AMD has announced that it is shipping a low power Mobile Athlon 64 3000+ processor for the thin and light crowd.  It is based on a low power 90 nm Athlon 64 (and I mean low power due to the way it was fabricated, in that power consumption was the focus and not speed), and it probably speaks well of AMD's 90 nm process to be able to produce a product like this.  Low heat and low power will enable longer battery life (the chip eats less, as well as any kind of cooling fan), which is key for many thin and lights.  This product could very well give the Intel Pentium M a run for its money in this market, but it is going to be a tough one for AMD to break into because it doesn't have an entire portfolio of products like Intel does with Centrino (chipset, graphics, wireless, and processor support).  While AMD may have wireless products, they certainly don't have their own chipset supporting mobile products (though the VIA K8M800 would be the first choice of many, and perhaps the SiS 760 may be another solid choice in the near future, as well as the upcoming ATI integrated product which will feature RV380 like technology for the graphics).

AMD is making a very solid push into the mobile market with the Athlon 64, and its features such as the NX bit as well as integrated memory controller and the simplicity of HyperTransport interconnects make it a very tempting product for laptop manufacturers.  eMachines made a very big splash with their Athlon 64 products, and they have been a retail top seller since they were introduced.  Compaq and HP have also released Athlon 64 based laptops, and apparently these are selling well also.  Intel still has a stranglehold on this market, but AMD appears to be chipping this away bit by bit.  The ability to produce thin and light Athlon 64's will help to sell more manufacturers on the features and benefits of the AMD product.  This truly is a watershed event for AMD and its Athlon 64 product line.

Acer should be one of the first out of the gates with a low power Athlon 64 notebook, and it will have all of the panache of the Ferrari branded products.  This will certainly catch more than a few eyes, and for people wanting good performance in a thin and light, as well as something different from the hundreds of other drab models featuring a black or grey exterior, this will be a compelling choice.  AMD is listing the price of each processor at $241, and it is available worldwide immediately.

As for other news... I know that other webmasters have said this as of late also, but I thought it was a big occasion.  For the first time since the release of the ATI Radeon 9700 Pro (2 years ago), I have another NVIDIA card in my main machine.  I recently received a MSI GeForce 6800 GT, and so far I have been very happy with it.  Unlike the GeForce AA6800 from Chaintech, the 2D is very sharp and clear (though I have heard from Ryan Shrout at PC Perspective that his model of the AA6800 doesn't have such a problem, but that the one from Tech Report does- I am thinking that the early 6800 OC product from Chaintech had some issues, but these apparently are cleared up now).  The performance of Doom 3 is a large step up from what my previous X800 Pro was able to do.  While I did enjoy the Catalyst drivers from ATI, I feel that NVIDIA has really caught back up and I like NVIDIA's setup a bit more.  The gaming profiles work very well, and the refresh rate override is set up really nicely and is easy to use.

The only problem with this MSI card is that it isn't the one shown on the website!  It does not feature the fancy copper heatsink, as well as the second power connector (it appears that particular 6800GT shares the same PCB as the Ultra product).  This product's PCB is red, and it has the regular NVIDIA single slot cooler (aluminum).  After I finish reviewing this product, I will be doing a few tests with overclocking while using a 6800 Ultra heatsink and fan on this GT board (donated by BFG for the cause).  Lots of content coming up to say the least!  Now if only I can get that Athlon 64 3700+ review finished!!!

September 17, 2004

AMD Delivers Desktop 90 nm Parts - Josh

AMD is finally catching up to Intel when it comes to shipping 90 nm parts.  Intel originally started shipping 90 nm Prescott based Pentium 4's to the market early this year, and only now has AMD caught up and released their own 90 nm parts.  Of course, the performance of the Athlon 64 on the 130 nm process from AMD has certainly allowed them a lot of leeway here, as they have not been forced to put out a 90 nm product to compete with Intel.  The first 90 nm desktop parts will not be the latest, fastest, and greatest though.  According to The Inquirer the first products to market are based on the Athlon 64 3000+, 3200+, and the Socket 939 3500+ SKU's.  These are definitely mid-range market products, but this is no big surprise coming from AMD.

The problem with 130 nm Athlon 64's is that they are very large in terms of die size.  A 512 K L2 cache Athlon 64 takes up a whooping 144 mm square on the 130 nm process, while the same (or nearly the same) product on the 90 nm process takes up only 84 mm square.  By doing some quick and dirty calculations, we can see that on a 200 mm wafer the 144 mm square product will allow around 180 dice per wafer.  By going to 90 nm and having each die take up 84 mm square will allow around 332 dice per wafer.  This allows AMD to nearly double the amount of dice per wafer, and from all indications their yields are very good for their 90 nm process.  The mid-range Athlon 64 products appear to be in great demand, and this is a good way for AMD to meet that demand.  Also, due to the design characteristics of the Athlon 64 at 90 nm, it is not nearly the power hog that the Prescott is.  Because AMD is not pushing their transistors to go as fast as possible, there is less leakage, less heat produced, and less power needed to achieve the 2.0 GHz and 2.2 GHz clock speeds that comprise the 90 nm products shipped.  This bit of information is supported by the motherboard industry, as AMD has not released updated specs for their products that increases either the power consumption or heat generation.  AMD has already been shipping the 90 nm Mobile Athlon 64's for about a month, and all indications point to those being some very cool running and energy efficient chips.

Due to the current competitive environment, AMD has had the luxury of sitting back and working on yields versus working on speed bins to keep up with Intel.  While the new Prescott based Pentium 4 from Intel is a solid chip, it does not add significant improvements in performance over what AMD currently ships, and in fact the top of the line AMD products outperform Intel in the majority of games and productivity apps (though Intel still holds a sizeable lead in things like video and MP3 ripping).  Intel is still having a heck of a time delivering large quantities of its top end 3.6 GHz Pentium 4, but the situation is greatly improved from when they first released it (you can actually buy this product online as well as from major OEMs).  AMD seemingly can produce as many 2.4 GHz Athlon 64 (1 MB and 512K) parts as they can, which is still their top of the line product.  The only thing that has held AMD back from significantly increasing their marketshare is the fact that these parts are still being made on the 130 nm process, which limits the amount of chips they can ship.  Now that the 90 nm process is picking up, AMD is going to be able to more adequately fill the demand for the high volume markets (those being served by the 3000+ through 3500+ processors).

Many feel that AMD will totally transition their Athlon 64/Opteron products to 90 nm by Q1 2005, and from all indications this looks to be the case.  The next speed increase will be Athlon 64 products on the 130 nm process, clocked at 2.6 GHz.  This really appears to be the upper limit of AMD's quite mature 130 nm SOI process.  By using their 130 nm process for this task, AMD can focus on getting good yields from their 90 nm process without having to struggle with getting good speed bins as well (as they did when they first transitioned their Athlon XP parts to 130 nm- we all remember the problems with the initial T-bread A parts).  This gives AMD around 5 months to further optimize their 90 nm process, as well as possibly create a revision of the 90 nm Athlon design.  5 months in this industry is a very long time, and the outstanding performance of the Athlon 64 design has really allowed AMD to get their ducks in a row.  Intel has had a tough time with their transition to Prescott and 90 nm, but they are slowly getting up to speed and by mid-Fall we can expect another round of product releases that again raises the performance bar.

September 15, 2004

AMD and PCI-Express - Josh

As many of you have probably read, The Inquirer has reported that AMD will not have a PCI-Express based solution out until early 2005.  This is in direct contrast to what I have heard.  I was told from a very good source that an AMD PCI-E solution would be available "very soon".  I do not interpret "very soon" to be "beginning of 2005", but rather when I was told this I was thinking the end of September.  Well, I still believe we will see the first solutions by October, which should probably give PCI-E a much needed boost in the minds of users.  While PCI-E is starting to take over in the OEM sector, it is not very high profile at the moment in the retail and enthusiast realms.  Currently AGP still rules, and the fact that the high end AMD processors don't have a PCI-E solution as of yet keeps it out of the limelight (since most sites agree that the AMD FX-53 processor on the socket 939 architecture is the fastest overall chip in the world).  So, when benchmarks are run, often they are done on the AMD platform.  If I am correct, and we see a PCI-E solution out in October, then we will probably see both ATI and NVIDIA really push the PCI-E 3D Graphics products out into the retail market (a very high profile market).

I did notice today that AMDZone had a very interesting link on their page to the Abit AX8 motherboard.  This little number is based on the VIA K8T890 chipset, which supports PCI-Express.  The board looks pretty finished to me, and the specifications do not look to be changed anytime soon.  So why would Abit put up such a product page?  Probably because it is darn near ready for release.  Many were expecting NVIDIA to be one of the first with a PCI-E product, but it looks to be later in the year before we see the next generation of nForce products.  SiS could actually be very close to having a solution with their SiS756 PCI-E product, but SiS has never been particularly popular with the motherboard manufacturers and their top of the line products.  This is unfortunate, as SiS makes some very solid products, but they are continually viewed as a "value" level provider.  In my opinion, I have found that their products are usually more solid, and perform slightly better, than the VIA products they compete with.  Also, it appears that SiS's SATA controller is a lot more solid than VIA's (which is actually just a bridged solution, and not fully integrated like NVIDIA's nForce 3 250 series).

NVIDIA has shown in the past that they can make a very solid chipset.  While the first generation nForce was not a huge success, it did pave the road for the nForce 2 (which proved to be the best overall Athlno based chipset for its generation).  The nForce 3 150 was a bit of a flop, but NVIDIA came back with the nForce 3 250 series, which is again proving to be a very successful part.  The next generation looks to improve upon that, and NVIDIA appears to not mind making it to market after its competition.  If this extra time spent equals a better part, then coming to market 3 months after the competition is not a bad thing, especially considering that due to the architecture of the AMD Athlon 64/Opteron, a solidly built next gen nForce will probably be in the market a very, very long time.  The nForce 2 has lasted a very long time, and I fully expect to see NVIDIA's PCI-E products to remain relatively unchanged for the next year and a half after its introduction (which measured in chipset time is quite a span).  If NVIDIA can produce a very solid PCI-E product, along with a robust feature set to go along with it, then they will again be in a dominant position in the AMD market.  While it has been essentially confirmed that there will not be a SoundStorm 2 functionality integrated into these next chipsets, do not discount the possible implementation of the Azalia technology standard.

So, we have two possible candidates for AMD and PCI-E, but it looks like the first retail boards will belong to the VIA crowd.  I am trying to get an official line from either VIA or Abit about this, and as soon as I hear something (that I can mention) then I will post it.  I seriously expect to have AMD/PCI-E solutions by October though.

September 13, 2004

What a Week! - Josh

This was the week where if something could go wrong, it did!  Hardware was breaking all over the place, nothing seemed to work, and every time I sat down to write for the site, something came up!  I am also in the middle of testing the high speed network D-Link sent me, but unfortunately the DI-624 router died in the middle of everything.  Not a whole lot of fun there!  I am also trying to set up an AMD interview, but the folks down in Austin are pretty busy at the moment.  It was awfully nice of George from Intel to take time out to answer my questions though, very kind of him!

On the technology front it seems the the keyword is "dual core".  AMD and Intel are both pushing their dual core strategies, and I am very curious to see Intel's answer to the dual core Opteron/Athlon 64 (the dual core Itanium is a bit beyond the market that the Opteron resides in).  Intel promises a full lineup of dual core products from the notebook to the server market, so we can expect to see a mix of Pentium M and Pentium 4 dual core products (though I am somewhat leary of a dual core Prescott, if one will ever exist).  Intel has really worked hard on HyperThreading support, so most productivity software has support out of the box for dual processor (as do some other non-productivity titles).  Dual core will really help AMD, as they currently have no HyperThreading type support, and in applications where both memory bandwidth and throughput are required (think video editing, ripping, etc.) having a dual core product will help to close the gap AMD has with Intel in this field.  I would be surprised to see if many more games started to become dual processor aware, as in those situations it isn't always easy to code for (just ask John Carmack).

Also of note is that NVIDIA appears to be supplying the market quite nicely with GeForce 6800 parts (from the vanilla, to the GT, to the Ultra).  I think that IBM has really solved many of its fabrication problems this summer, and the latest revision of the NV40 chip is really coming around.  ATI is in a strange place with its X800 XT PE series of cards, as they are really not available in the retail market.  It seems OEM's are getting the lion's share of these products, which means that ATI is not really meeting demand for this high end product.  Is there trouble brewing with this design and TSMC's 130 nm Low-K line?  I am not sure, but I bet that the 130 nm L-K line is running at full bore, and TSMC just can't produce enough chips for ATI (as several ATI products utilize that technology).

Well, that is all for now.  As soon as I can get anything else up, you will be the first to know!

Copyright 1999-2004 PenStar Systems, LLC.