: News
: Reviews
: Editorials

: Archives
: About
: Contact
: Advertising
: Privacy
: Links
: Forums

 

 

 

 

 

 

 

 

 

December 18, 2007

Hardware Availability

I am happy to see that the HD 3870s, AMD 790FX, and G92 8800 GTS products all have good availability around the web.  The 3870s are starting to fall down in price, with many models hitting the $249 level while the lower end 3850 is right where it should be at $169 to $179.  512 MB models of the 3850 cards are also quite appealing at $209 and below.  Most of the big manufacturers have been able to meet demand for their high end AMD 790FX boards, and there is a good selection with prices starting around $179 up to $269.  The new 8800 GTS is also well represented, with products at price points around $329 after rebate.  Not a bad deal considering most of these cards come with a title like Crysis or CoD4.

So what is the bad news?  The 8800 GT is still the hot ticket this Holiday Season, and supplies are pretty tight for this midrange wonder.  The new GTS has taken a bunch of the G92 chips that are also used in the 8800 GT, but the extra $50 to $75 that the jump entails has made the GTS rather unappealing to many in the face of the 8800 GT.  There is decent availability of the 256 MB model of the 8800 GT, but the 512 MB version is what most people seem to want.  This may be a small blessing for AMD and its partners, as the HD 3870 and 3850 cards are all well represented and at competitive price points.  While brand loyalty is pretty strong, the temptation to buy a card at this time of year could push some consumers over to the red side.

Memory prices are hitting an all time low, and a user can pick up 4 GB of DDR-2 800 with slightly relaxed timings for around $100.  That is just sorta insane.  I love it though.  Upgrade to some good CAS 4 memory, and a consumer is paying around $135 to $150.  As mentioned in my Super Talent interview, most manufacturers are expecting memory prices to firm up at the beginning of next year, so it might be in your best interest to make that memory upgrade while prices are this dirt cheap.

Be sure to check my mini-editorial about the Puzzling Existence of the AMD 770 Chipset.  Its a good one.  Or at least I pretend it is.  Leave a comment if you dare!

December 17, 2007

NVIDIA Releases the nForce 780i

The nForce 680i has been a staple in the enthusiast category of chipsets for over a year now.  This piece of tech has been remarkably long lived, and quite successful.  While a few problems had to be ironed out throughout its lifetime, it has been the basis for quite a few high end machines from both DIYers and OEMs alike.  It provided excellent performance combined with good overclocking potential.  NVIDIA's new 780i is an interesting beast which is more evolutionary rather than revolutionary though.

I think the best way to categorize this product is "680i on a bit of HGH".  There are three chips that make up the 780i platform, and two of them have been seen before.  The northbridge is actually comprised of two chips, the 780i SPP and the nForce 200 PCI-E Gen 2.  The 780i SPP is functionally identical to the older 680i, but likely it is a slightly tweaked and respun version.  Its 16X PCI-E link is "overclocked" and used to connect to the nForce 200 chip.  It is clocked up to 4.5 GT/sec, rather than the fully spec'd 5 GT/sec that PCI-E Gen 2 supports.  Theoretically it should give more than enough bandwidth to the cards sitting on the nForce 200.  The nForce 200 is a new chip that supports 32 lanes of PCI-E Gen 2.  The southbridge is actually the nForce 570 SLI chip, which connects to the northbridge through Hypertransport and has another 16 lanes of Gen 1 (plus a 12 extra PCI-E lanes going to other slots if needed).

So basically we have a mainly recycled platform with some extra bug fixes and support for PCI-E Gen 2 and Tri-SLI.  It also fixes the incompatibilities that the 680i designs suffered with Intel's latest 45 nm processors.  The products being released today do not officially support the upcoming 1600 MHz FSB parts from Intel either.  NVIDIA has stated that they will have a new part ready for when those processors actually hit the market... which looks to be a late January to March timeframe, depending on how Intel wants to handle its own little "errata" problem with their 45 nm quad parts.  The chipset also supports memory speeds up to DDR-2 800 speeds, though NVIDIA claims that the memory controller works perfectly fine up to 1200 MHz speeds.

The new boards support the latest version of nTune, as well as NVIDIA's new ESA platform.  NVIDIA should be releasing these programs today.  Overclocking support out of the bag appears good, and they are doing their best to cater to the enthusiast market.  The 570 SLI chip is actually quite robust in terms of features.  Dual Gig-E, 6 SATA 2.0, ATA channels, 10 USB ports, and High Definition Audio.  It does not support AHCI, but with the nForce drivers installed it does support hot swappable drives and NCQ.

The biggest problem facing the 780i is power consumption.  The older 680i sucked up quite a bit of juice, and adding the nForce 200 to the mix will not help.  Sure enough, it appears as though the 780i will lead the pack when it comes to eating up your energy dollars.

The 780i does look to be a decent chipset, and quite a few manufacturers will release parts based on both the reference design (XFX, EVGA, BFG) as well as original designs (Asus, MSI, Gigabyte, etc.).  This will help plug the gap for a while between NVIDIA's offerings and the other PCI-E 2.0 compliant boards out there (based on Intel's X38 and AMD's 7 series).  NVIDIA is working on another product that will address the PCI-E 2.0/1600 MHz FSB space, but won't be delivering this for a couple of months.  Until that time, the 780i will be NVIDIA's top end offering, as well as being the only one support full speed Tri-SLI (the 680i supports it, but one slot is not full 16X speed).

Tech Report has a nice little overview of this number up, and one should go read it to get a better idea what I am talking about.

December 14, 2007

AMD's December 2007 Analyst Meeting

Yesterday AMD hosted a meeting in New York for analysts and members of the press.  This meeting was of great importance for AMD leading into 2008, as their record of 2007 was not exactly shiny.  Some of the "lowlights" of 2007 were the continued loss of money due to the ATI acquisition, staying behind NVIDIA in the graphics performance race (as well as losing marketshare throughout the first half of the year), inability to scale their 65 nm process to where they needed it to go, and the big one being the inability to ship fully functional quad cores to market in both desktop and server.  While that makes for a pretty negative bundle, not everything is looking down for AMD.

Right now the shining star for AMD is the RV670 chip powering the HD 3850 and HD 3870 graphics cards.  The 3850s are apparently selling quite well, and AMD pushed a lot of chips towards that SKU.  Unfortunately AMD underestimated demand for the higher end 3870 model, and did not ship nearly enough of that product to their partners (or their partners underestimated demand for those cards and didn't order enough high end chips).  Until the past few days, availability of 3870s were poor.  AMD is currently shipping a larger portion of high end chips to their partners, and we are seeing better availability.  It is unfortunate that they could not address the demand a little sooner, as the Holiday Season is well upon us.  It seems the partners have also tried to help the situation by releasing 512 MB versions of the 3850, and that apparently has helped quite a bit, as those are turning out to be popular SKUs at their price points.

AMD's 7 series chipsets are starting to get a lot more attention as well.  The first value 770 based boards hit the market from ECS and Asus, and several other manufacturers are about to release their boards.  There are quite a few 790X and 790FX boards at pricepoints over $150, and availability of models such as MSI's K9A2 Platinum have been tight.  Apparently this is a very popular board due to its relatively modest $179 price, as compared to the much more expensive Gigabyte and Asus models now available.  The 690G made AMD a lot of inroads into the integrated sector, and there is a lot of interest in the upcoming RS780 integrated part.  The RS780 is a DX 10.1 compliant part, and is said to be around 4X more powerful than the previous 690G (which is the fastest current integrated part).  RS780 was shown in New York to select journalists, and its "gotcha" feature is the ability to perform Hybrid CrossFire with the upcoming HD 3450 (RV620 based).  This $49 add-in card is competent in 3D graphics, but enabling Hybrid CrossFire improved framerates by around 50%.  Ryan over at PC Per (my other haunt) has a good write-up on what he saw.  While Hybrid CrossFire will not be for everyone, it does open up a lot of possibilities.  As a standalone part, the RS780 does look to be a pretty powerful integrated part which may hold the integrated performance crown throughout 2008.  NVIDIA may have a competitive DX10 part, but until this point very little has been heard about it.

Something of interest that was skipped over was that the test board they showed looked like it also had the upcoming SB700 southbridge.  This chip was originally supposed to launch with the other 7 series northbridges, but delays have kept it to the early 2008 timeframe.  The current SB600 chip is chugging along, but it is not the best solution out there for heavy I/O tasks that are typically associated with southbridges.  The SB700 is supposed to fix a lot of those deficiencies plus add a few more features to the mix (namely more USB and SATA connections, better I/O performance, better support for hybrid drives and flash).

With the RV670 chip, AMD is continuing to address the upper graphics market.  While prices for the HD 3870 have not yet gotten back down to the suggested MSRP of $219, they are slowly making their way.  In January AMD will be releasing their 2 x RV670 card (officially considered R680) as their high end variant to compete directly with NVIDIA's 8800 GTX/Ultra.  AMD has already shown this card at multiple events.  Its power draw is said to be comparable to the older HD 2900 XT.  AMD has not abandoned the high end of the graphics market, and if their CrossFire support continues to improve it should be a pretty good card for the money.

The one thing that AMD really pushed at the analyst meeting is that things are not as grim as they seem.  It has more Opteron customers and more commercial client platform partners then it ever has before.  With the acquisition of ATI, they are also shipping a much wider selection of products to markets other than the PC and Server space.  AMD really is no longer a one trick pony.

AMD will not have a break-even quarter as they were expecting in Q4, but guidance has been revised to show that they expect Q2 2008 to be the breakeven point, and Q3 should show a profit.  Of course, take that with a grain of salt as a lot of things can happen in that time.  AMD has yet to produce good volumes of fixed Phenom (Rev. B3) cores, which is their bread and butter for both desktop and server markets.  AMD is revamping the X2 lineup and will release eleven new 65 nm products starting in 2008 to fill the gap left by the late arrival of Phenom and its derivatives.  From all indications, AMD is finally on its way to getting their 65 nm process in line with what it should be.  This situation is actually quite reminiscent of AMD's problems with their 250 nm process and the K6 architecture.  Just as then, AMD will eventually fix the issues surrounding their process.  AMD is very bullish on their 45 nm transition, and are expecting to ramp production in the first half of next year, with volume availability of 45 nm parts in early 2H 2008.  It is an aggressive ramp, but AMD's partnerships with IBM and Chartered may in fact have paid off.  Hopefully for consumers this is the case and we will see higher speed 45 nm quad and dual core parts with IPC improvements over the current 65 nm X2 and Phenom products.

The company does not appear to be circling the drain, and the company has indicated that much of the extreme bleeding of money has stopped.  They will announce the extent of their "material goodwill charge" on the ATI acquisition, which basically means that AMD paid too much for that company.  If AMD can show that they have surpassed most of the major issues they had in 2007, then 2008 could be a good year and they could be well on the way to turning around the boat.  They have some aggressive platform plans for the future, and one area of strong growth for them looks to be notebooks.  The upcoming "Griffin" processor (mobile X2 with power optimized memory controllers and HT 3.0 connections) and the RS780 chipset, when combined is known as the "Puma" platform, may prove to be a very appealing product.  Hopefully it will make the company some money.

AMD has a big year ahead of them, and the company is indicating that they are well prepared for it even after all of the missteps of the past two years.  They will have competitive products at reasonable pricepoints, and their new platform mentality could very well pay off in several important markets.  If AMD can in fact deliver on the fixed Phenoms in a reasonable timeframe, then things will be looking up for the company.  Even if they move to Q2 for volume Phenom release, they still have a lot of products holding the company up.  Needless to say, 2008 is going to be an interesting year for AMD.

December 12, 2007

8800 GTS (G92) Release and Other News

Yesterday NVIDIA and its partners introduced the new (and mostly improved) GeForce 8800 GTS.  This is the 65 nm based G92 product, clocked in at 650 MHz core, 1625 MHz shader clock, and memory pumping along at 970 MHz (1940 MHz effective).  Though it is 4 ROPS shy of the old 8800 GTS, it makes up for it with a higher clock speed and a full 128 stream processors vs. the old one's 96 (clocked at a relatively piddling 1200 MHz).  Memory bandwidth is about the same between the two, and considering that the lower clocked 8800 GT beats out the old GTS by quite a margin, we know that the new GTS is trading blows with the more expensive 8800 GTX.  As listed in my 8800 GT article though, we know that the GTX still rules the roost when high resolutions and AA are enabled, but for 1920 x 1200 and below gaming... the new 8800 GTS is hard to beat.

Prices are sitting above the $349 level that NVIDIA indicated would be the top end, and the lowest priced unit I have seen so far is $360.  This is not exactly unexpected, as prices on the 8800 GT are still well above the $199 to $249 that NVIDIA suggested.  The higher price of this product, along with the competition for consumer's dollars that the 8800 GT represents, has allowed the 8800 GTS to stay in stock a day after release.  A far cry from what the 8800 GT did (which was sold out at most places for about 3 weeks unless a consumer got lucky and caught a shipment coming in).

The dual slot cooler does a much nicer job of keeping heat under control on the GTS, which again is unlike the 8800 GT.  It is also whisper quiet all of the time.  The card does not support the upcoming Triple SLI, but it can still be used in SLI.  The GTX and Ultra are still the only cards that can do the "trip".  The card is the same length as the 8800 GT, which makes it a nicer product to shove into tight cases as compared to the GTX and Ultra.  Power consumption is also down by a significant margin.  The only real issue I have with the GTS is that it "only" has 512 MB of memory.  Most applications fit nicely into that local memory, but we are starting to see a few games which really push that limit.  Especially when AA is enabled.  Still, it is another nice option for a video card upgrade, and we are happy that these are available for the holidays!  You can catch a few reviews at some of the regular places:

PC Perspective

Tech Report

Anand's

HardOCP

In other news I am starting to do a bit of work for PC Perspective.  I am primarily writing some editorial content, doing some news, and helping with the podcasts there.  I am still making this my primary home, but you gotta pay the bills you know.  Reviews and editorial content will still be making it up here, but overall I am aiming at being more productive at both places.

December 10, 2007

NVIDIA's GeForce 8800 GT

Last month NVIDIA dropped the first "must have" computer component of the Holiday Season, the GeForce 8800 GT.  This 512 MB card was based on the new 65 nm G92 chip, which encompasses all of the features of the G80 and G84/86 chips, plus adds a great amount of power savings to boot.  NVIDIA was kind enough to send me a reference board to take it around the block (or two).  Here is a quick quote from the review:

            Before heading into the specifics of the card, I thought I would talk a bit about the G92 chip that this card is based on.  The G92 is fabricated by both TSMC and UMC on their respective 65 nm processes.  The 8800 GT features a G92 chip that has portions of it disabled.  The G92 chip features 128 stream processors, 64 texture address units, 64 texture filtering units, and 16 ROPS/Multi-Sample units.  In the 8800 GT one shader cluster with its associated texturing cluster is disabled.  So the 8800 GT features 112 stream proessors, 56 texture addressing and filtering units, but keeps the 16 ROPS.

You can read the entire article here.

December 3, 2007

Interview with Joe James of Super Talent

Being the rather inquisitive person I am, I thought it a good idea to ship off a few questions concerning Super Talent and the memory industry in general.  Joe James was kind enough to to respond.  Here is a quick example of the interview

How long has Super Talent been around, and when did they decide to pursue a more retail oriented business plan?

Super Talent was founded in 1991. The company historically kept a low profile, making memory for private label OEMs, and quietly became one of the world’s top memory manufacturers. A couple of years ago our management decided to transform the company – to refocus the company’s vast technical and manufacturing resources on developing leading edge products and building our own brand. We’ve worked hard for the past two years to become a leader in SSDs, server memory and extreme performance DDR2 and DDR3.

You can read the entire article here.

November 26, 2007

AMD Athlon X2 5000+ Black Box Edition Review

From the launch of the original Athlon 64 3200+, AMD has been the choice of enthusiasts everywhere.  Well, until the Core 2 Duo showed up and ruined the party.  Since that thrashing AMD has been trying to woo the enthusiasts back, and one of their recent rather successful ventures has been the introduction of the Black Box processors.  These unlocked processors are aimed squarely at the enthusiast, and AMD is certainly hoping to give a little back to that community that supported it so well.  Unlike the X2 6400+ version, the 5000+ is made on AMD's 65 nm process.  The combination of unlocked multiplier and 65 nm performance could be enough to turn a few heads.  Combine this processor with a new 790FX board, and the average user becomes far less average.  AMD's very own Herbal Viagra if you will.

            All was not wrong with the 65 nm X2 though.  While the design was slightly slower than the older version, it was significantly more power efficient.  AMD rates the 65 nm X2 as a 65 watt product across the board, except in the case of the X2-BE products which are rated at 45 watts.  The older X2s have seemingly been sitting pretty at 89 watts for ages.  This little marketing point became the mantra for the company, as energy efficiency trumps all these days rather than performance at all costs.  Just ask Al Gore.

You can read the entire review here.

November 20, 2007

AMD's Spider Platform Launch

Last week AMD was kind enough to invite me to Lake Tahoe to view their Spider Platform launch.  The Spider is comprised of the newly released Phenom processor, HD 3870 video card, and the 790FX chipset.  Originally we were very excited about the prospect, until we learned that all benchmarking was to take place there.  Yes, no Phenom for the home lab.  Not to be deterred, I thought I could get some of the deeper stories behind this launch rather than just the benchmarks from the unreleased 2.6 GHz part AMD provided the attendees.  In chatting with the AMD luminaries in attendance, I received a deeper understanding of the technology and decisions behind this launch.

            Phenom was supposed to herald AMD’s second coming (or was that the third?).  A native quad core processor with an integrated memory controller, Hyper Transport 3.0 connectivity, and a host of enhancements that would increase IPC, as well as a split power plane design which would improve overall power efficiency and scaling.  All of these aspects combined were supposed to make the Phenom more than a match for the mighty Core 2 Duo and Core 2 Quad processors from Intel, both 65 nm and 45 nm versions!

You can read the entire experience here.

November 19, 2007

Today's AMD and Intel Launches

Last week I was invited to Lake Tahoe by AMD to get hand's on experience with their Spider Platform.  For those not in the know, the Spider Platform is comprised of the AMD Phenom processor, the Radeon 3850/3870 graphics cards, and the 790FX/X chipset.  AMD had high expectations for this launch, and I would say that two out of three of the products do live up to expectations.  The graphics cards that AMD released last week are simply top of their class when looking at price points and power consumption.  The Radeon 3850 overpowers the HD 2600 XT and GeForce 8600 GTS at their pricepoints, and the Radeon 3870 undercuts the GeForce 8800 GT by up to $70, yet delivers performance that almost catches up to it.  AMD had pretty good quantities of product at last week's launch, but the top end cards have sold out everywhere, and the 3850 is in limited supply as well.  We are expecting another big batch of cards to be delivered to retailers this week though, and the prevailing rumor is that AMD will have provided over 500,000 cards to the marketplace by the end of this year.

The disappointment to this launch is the availability and performance of the Phenom desktop processor.  Originally AMD had expected to launch at speeds ranging from 2.2 GHz to 2.6 GHz.  Things did not work out for a variety of reasons in this area, and AMD is only launching the 2.2 and 2.3 GHz Phenoms (model numbers 9500 and 9600 respectively).  Apparently an errata with the Translation Lookaside Buffer for the L3 cache exposes itself at 2.4 GHz and above.  In certain situations this can cause an error and will hard lock an application or the OS.  AMD has a new revision chip in the works that corrects this problem, as well as a couple of other speed issues.  There has been some conflicting information as to when this new revision will hit the marketplace, but I am pretty certain that we will not see this new chip until January at the earliest.

I am finishing my complete article on the experience and the technology, and will have that up later today.

In the meantime, Intel has decided to further spoil AMD's party by seeding reviewers with the QX9770 processor.  This 3.2 GHz quad core monster runs on the yet unofficial 1600 MHz FSB (400 MHz quad pumped).  Intel's X48 chipset will be the first official product to support this bus speed, but most X38 motherboards have no problem running at this speed (though of course, unofficially).  The processor is simply a barn burner.  Utilizing the 45 nm Penryn core, Intel lays the smack down on AMD in performance.  This processor will not be available until January, 2008.  When it does hit the streets, expect a $1200+ price tag.  It may not be entirely obvious to everyone, but Intel just did their best to spoil AMD's Spider launch.  I am not sure Intel really needed to do this though, as the Phenom 2.4 GHz has an awfully hard time keeping up with the Q6600 (2.4 GHz quad) in most applications.  The base Penryn core is an impressive product, and I am sure Intel will do very well with it until Nehalem comes out in late 2008.

All is not well in the Intel camp though, but this particular issue is not the fault of Intel.  The NVIDIA 680i SLI chipset does not play well with quad core Penryn processors.  Massive errors, failure to boot, or just simply turning off two of the four cores are the results of using a Penryn based quad on the 680i SLI chipset.  This has turned out to be a pretty major problem for NVIDIA, and one that has caused them to delay the arrival of their 780i SLI chipset (which is comprised of a 680i northbridge and a PCI-E 2.0 "adapter" chip).

AMD certainly has its work cut out for it.  Phenom gets AMD close to overall performance of the current Core 2 architecture, but they need to further optimize performance and get clock speeds up.  Oh yeah, and let's not forget actually getting product into the marketplace.

November 9, 2007

AMD 790FX Hits the States

While several online retailers have listed the Gigabyte based 790FX motherboard, it seems that Newegg actually has them in stock and are shipping them out.  Fully 10 days before AMD officially announces these parts, we are seeing these motherboards start to trickle in.  In other parts of the world, these boards have been available for several weeks.

Gigabyte appears to be the first out the blocks, and we can expect MSI and Asus to follow shortly with production boards.  For those of us living in the States, we can finally get our paws on the top end offering from Gigabyte.  At $269 it isn't exactly cheap, but at the same time it is the top of the line product from Gigabyte supporting the 790FX chipset.  MSI will likely be cheaper when they come to market, but we can safely bet that the top Asus board will be more expensive.

November 8, 2007

AMD Firestream 9170

AMD is announcing the Q1 2008 availability of the first double precision stream computing solution.  The chip being used in these Firestream cards appears to be the RV670, but with double precision float support in the stream processors.  On the upcoming desktop based cards, the double precision functionality is likely to be disabled.

The boards will pull less than 150 watts, but feature 2 GB of memory.  Other details such as the memory bandwidth or bus width are left out of the announcement.  It is much more likely at this time that the RV670 has a 256 bit path rather than the larger 512 bit memory bus that the HD 2900 XT/R600 does.

NVIDIA was originally supposed to release a Tesla product with double precision this year, but that looks to be delayed until near the end of 1H 2008.  This could give AMD a nice little window of opportunity.  Considering that each card will approach $2000 US in price, AMD could certainly use to have a high margin product for a change.  AMD also looks ahead of the game in terms of software support and their CTM language (Close To the Metal).  Historically speaking, we have seen more stream computing application support for the ATI/AMD cards first.

AMD does not give the rate at which double precision is computed, but rather keeps their figures in the 500 GFLOP range using single precision.  My best guess is that double precision will probably be ½ that (considering that the R600 was single precision and made up of 700 million transistors, and the RV670 has the same number of stream units but is 666 million transistors).  Likely there is some loopback mechanism that allows these units to support double precision (very much unconfirmed).

Still, it is a good announcement from AMD, and one that is needed to help get stream computing off the ground.  The biggest complaint so far for these products was the lack of double precision for scientific applications, but that has obviously been addressed by this release.  These cards will be much more welcome in servers doing this type of work due to the lower power requirements, and the lower amounts of heat. 

GeForce LAN 4

November 16 through 18 NVIDIA is holding its latest GeForce LAN.  The event will be held in Alameda, CA on the old Alameda Naval Airstation (Area 51 Event Center).  Crysis will be the big attraction for this LAN, and I am sure there will be a few more surprises at the LAN itself.  Only 450 spots are available for purchase in the BYOC LAN, so those who will be able to attend should sign up and purchase their spot ASAP.

There look to be some very good sponsors, and over $60,000 worth of hardware and software will be given away.  Likely each BYOC attendee will be receiving a full copy of Crysis courtesy of EA.

8800 GT Selling Like Mad

In what is likely not to be a shock, the selection of available 8800 GT cards is quite slim.  Newegg continuously gets in stock, but sells out within hours.  Other online retailers show much the same story.  Prices get down to around $269 for available parts, and stretch well into the $320 territory.  Once demand settles down and retailers get a better stream of parts, expect prices to start dropping to the $200 level for the vanilla offerings, while overclocked parts with plenty of extras will fall well below $300.  Or so that’s the plan.

The 8800 GT has really revitalized the upper midrange market, and people who were hesitant about spending $250 to $300 for 8800 GTS 320 parts are seemingly jumping at the opportunity to get the 8800 GT with its 512 MB of memory.  Performance for the stock parts are just shy of 8800 GTX levels, and more often then not are significantly faster than the older 8800 GTS 640/320 cards.  NVIDIA certainly hit a homerun with this release, and users are quite happy to get that kind of power in hand for that price (not to mention smaller size and single slot cooling).

November 7, 2007

NVIDIA's ESA

A few days ago I was invited to view a presentation by NVIDIA concerning their plan to introduce a comprehensive monitoring and tweaking system for the personal computer.  The Enthusiast System Architecture is an open source (sorta) endeavor to utilize a standardized communications method to monitor and adjust system settings.  The system uses USB to communicate from the peripherals to the OS.  The plan is to make the system a lot more responsive to its power and thermal needs, as well as optimize its behavior under different loads.

While the peripheral manufacturers (power supplies, cases, water cooling equipment, etc.) are free to utilize this, I am guessing that initially only NVIDIA nForce products will be able to utilize the front end application that runs it all.  NVIDIA has produced a nifty looking program that works through OpenGL to display all the relevant information, as well as make adjustments to manually improve things like fan speed, water flow, etc.

If the USB-if (USB governing body) accepts this specification, then it will be open and free to use for all manufacturers.  If it is a success, I would bet that AMD and Intel would also create custom front ends for their chipset based products to utilize ESA.

The main idea behind this seems to be getting the ball rolling on a more dynamic system architecture.  A system that can modify and monitor itself, optimizing power usage and thermals, all the while giving the user the performance they need depending on the application.  If a user is playing a heavy duty 3D game, the system gives the maximum performance and increases the cooling needed, if the user is playing solitaire then the power requirements are lowered and cooling is turned down to near silent.

It is a good idea, and one I do hope the industry will embrace.  The use of USB should make it easy to implement, and considering the amount of internal USB headers that often go unused on motherboards, this is a good place to start.  Products will be made to help reduce the potential cable clutter that it will entail, but for now we will have to deal with thicker cables snaking throughout the interior of the case.  The idea is good, the implementation appears simple, and the benefits for the user are there.  Now we just need to see if it is taken up by other manufacturers.

November 6, 2007

More AMD Chipset Announcements

Asus and Biostar are two of the latest to officially release AMD 7x0 chipset based motherboards.  As I had mentioned before, we will be seeing many more announcements before AMD officially releases these new products.

Asus features two boards based on the AMD 790FX chipset, which is the high end offering.  These models will likely be well above $175, with the top end getting close to $250.  AMD does not appear to charge much for their chips unlike the competition, and I am not entirely sure if that is a good or bad thing.  In the previous generation, the 590 SLI boards were always at least $50 more expensive than the corresponding AMD 580X models.  I am pretty sure that both of the AMD 790FX based boards will be less expensive than the other top Asus X38 and CrossFire products.  Something of great interest here is the specification of DDR-2 1066 memory being available with only certain CPUs.  The meaning here is that Phenom will launch with 1066 support.

Biostar is launching a slightly different product, but one much in the same vein as the old TForce 550.  Using the lower end AMD 770 chipset, this budget oriented TA770 A2 looks to be the first officially announced AMD 770 board.  Considering how well the new AMD chipsets are supposed to overclock, this budget board could be yet another gem from Biostar for the budget enthusiast.  The board is not feature packed, but it does follow along with their older and current models in providing a good overclocking experience for not a whole lot of scratch.  One pleasant and surprising thing is that the onboard GbE port is powered by Marvell rather than Realtek.  The overall design looks pleasantly roomy, not to mention logically laid out.  This could be a really nice board for those not looking to invest in CrossFire.

So far little has been heard from the new SB700 southbridge, and most of these designs use the older SB600 chip.  Due to the pretty hefty number of PCI-E lanes available, most manufacturers are fitting their higher end boards with plenty of other chip options, such as dual ethernet, E-SATA, etc.  I honestly can't wait to get my hands on one of these 7x0 boards to see how well they in fact overclock.  Fun times ahead!

November 4, 2007

Updates to The State of 3D

As always, after I post one of these types of articles, more information comes out of the woodwork.  So here is a small update for those interested.

It may not look like the 8800 GTS 512 (G92 based) has a 320 bit, or 384 bit for that matter, memory bus.  NVIDIA had a interesting blurb on this page, but now it appears to be taken down.  Basically, if the G92 had a memory bus greater than 256 bits, then the memory density would reflect that.  We would then expect the boards to have 640 MB rather than the stated 512 MB.  As of right now though, NVIDIA has corrected their mistake and taken down all reference to the GeForce 8800 GTS 512.  Considering that NVIDIA is holding the GeForce 4 LAN in the middle of this month, we can assume when to see the new 8800 GTS 512 come out.

NVIDIA is doing a refresh on the older G80 8800 GTS, and EVGA and BFG Tech have already announced their products.  Basically it is a full G80 8800 GTS with 5 ROP clusters, 112 stream processors, 7 texture address/filter clusters, and the full 320 bit memory bus.  The core speeds for these products are now around 570 MHz (depending on the model), with slightly faster memory.  The SPs also were given a speed increase.  Overall, this particular card should compare well with the new 8800 GT 512, as well as the older 8800 GTX.  It is an interesting SKU because it does feature the larger memory bus, and greater pixel fillrate.  While the 8800 GT might be faster at lower resolutions, once things get cranked up with AA added, then the "new" 8800 GTS 640 should pull ahead.

On the AMD side, I think I was a little too optimistic about the performance of the RV670.  I honestly was expecting a "miracle chip" from AMD, but it appears as though the RV670 will be about as fast as the R600 based HD 2900 XT, but it will have a higher number and support UVD.  From all indications, the transistor count of the RV670 is lower than the original R600, and much lower than my inflated speculation of nearly 800 million transistors.  Indications are that it is comprised of around 650 million to 680 million transistors.  When you consider that UVD was added as well as other things, we can probably guess that the overall chip is not going to be all that much faster than the current HD 2900 XT.

While this has not been confirmed so far, I am afraid that AMD will not take back the midrange crown from the NVIDIA GeForce 8800 GT.  We will find out around the week of Nov. 15 whether AMD has a good contender on their hands, or if they will have to lower their prices on these boards to keep competitive.  I am honestly not too optimistic

I am still looking forward to AMD's new chipsets though, and the boards featuring these new chips are already out in the market (though not in the US quite yet).  The 790FX looks to be a neat product, and while not as feature packed as NVIDIA's chipsets, it does appear to pull a whole lot less power.  Considering that is one of AMD's primary concerns with their latest products, it should fit in quite nicely.

A Fun Place to Visit

The other day I joined some fellows over in Cheyenne for a bit of LAN gaming.  Point Blank Gaming is a small group of guys who like to get together now and then and shoot things up (virtually).  They also have some pretty active members on the message boards, and a variety of topics are often covered.  Some good folks, and I visit there pretty much every day.  If you are in the area, and are interested in checking it out, drop by and say hello.

Their next big LAN should be held this coming Spring.  The last one was just before Halloween, so they probably need the break after the whooping that I gave them.  Just kidding.

 

 

If you have found this article interesting or a great help, please donate to this site.

 

Copyright 1999-2007 PenStar Systems, LLC.