iXBT Labs - Computer Hardware in Detail






XGI: Legends and Myths

It seemed that the slack start of Volari series products, which were actually soundly defeated on the market, would break XGI. But no, it's not so fatal. The company is alive, it demonstrates a card based on the new XG 47 chip at the last CeBIT, and shares its multiple plans for the future. Well, let's analyze XGI plans and announcements and try to work out which of them have chances to be implemented and probably go down in the legends, and which of them risk remaining marketing and technical myths.

Legend One: Company strategy


So, the Volari series is intended for the market with the highest potential (Mainstream and Value sectors). The question who gets what share of this potential is not raised on this slide. But judging from small (in size) competing product titles, XGI has views on this market sector, being the main one for discrete graphics. It's also noted that the sector of high-performance products is a small sector. That's actually true, especially judging from the number of sold chips. But on the other hand, we shouldn't forget that the margin (profit) from one accelerator manufactured is several times higher. And the fact that now is when this sector influences the brand forming-up and customers' opinions of the brand most. But these considerations cannot be applied to XGI due to physical impossibility (at this moment) to design and manufacture a product complying with high standards of modern top accelerators from ATI and NVIDIA rather than due to the strategy (described on this picture). Well, let's forget about top solutions and see what XGI can manufacture and offer in middle and low ends.

Legend Two: Roadmap


According to this legend, which looks more like a myth so far, in 2004 (remember: it's the last year) there must have appeared (or at least designed and ready for production) two chips – XG47 and XG45 (offers higher performance). Judging from the XG47-based video card demonstrated at CeBIT, this chip saw the light exactly at the end of 2004. But the mythic XG45 is known only from the archeological side, that is only from such diagrams.

So, both chips are announced to be based on 130nm technology and PCI-E bus. The second point raises some doubts – everyone knows that the PCI-E sector is currently overcrowded as it is, and the main sales fall on off-the-shelf full configurations, where ATI and NVIDIA's positions are in peak strength. A single user may risk an experiment and buy an XGI to upgrade the old accelerator, but system integrators take into account many issues different from performance and price when they buy components wholesale and they will hardly choose Volari products (no matter how badly we want it). But on the other hand, from the mythology point of view, this bus choice looks justified – this product is advanced, trendy, and even revolutional for XGI.

Keep mysterious XG45 in mind. But now let's take a look forward along the chronological axis. In 2005 we'll meet (considering the experience of the previous generation, 2005 must be read as 2006) two new chips manufactured by the 90nm process technology. To all appearances they are architecturally almost the same, but in a year (in 2006, that is actually in 2007), we'll see no less than the WGF 2.0 platform codenamed XG6X. It will feature PCI-E II as well as compatibility with Longhorn graphics and driver standards, probably including Shader Model 4.0. Sounds legendary, if (glory to investors) the company and its product line live till 2007. It's quite convincing though that such tempting words as WGF 2.0 and PCI-E II should soften investors' hearts, long may they live, and that XGI will at minimum get an opportunity to raise the level of its chip design skills. Right up to products that can really offer every bit of a competition and conquer a noticeable share of the market even without additional optical devices – that's if this fairy tale has a happy end.

Legend Three: XG47 specifics

Key features of XG47:

  • 130 nm.
  • DirectX 9.0c
  • Shaders 2.0
  • 4 pixel pipelines
  • 64bit memory bus (DDR)
  • RUMA technology
    (a counterpart of ATI HyperMemory and NVIDIA TurboCache)
  • Dedicated hardware video processor
    (deinterlacing, motion compensation, IDCT)
  • DVI (1600x1200), HDTV (1080i/720p) support
  • TV - Out and LCD panel interfaces built into the chip
  • There may be fanless solutions, low power consumption (mobile solutions).

Interestingly, the company materials demonstrate an attempt to position this solution for media PCs, especially compact and inexpensive. That's right – good video features, no active cooling, using system memory via PCI-E. It's good for inexpensive monoblock units connected to TV sets, this market will grow in the nearest future (its growth will depend in many respects on the success of HDTV mass sales, if it happens at all). However media PCs are rather expensive so far and customers often need sufficient gaming performance (if it's a concept of an all-purpose entertainment center, it should be used in full, as a game box replacement as well), not only for playing video and audio, which can easily be done using a much less expensive DVD player.

At first, RUMA details:

We can see nothing innovative here, except for a daring configuration completely without local memory. Theoretically, it's quite possible and viable - in case of absolutely no 3D tasks – that is when a PC is used for video content, internet browsing, and office applications only. But an attentive reader may recall the chipset-integrated graphics, which can manage this task well and sometimes even provides fair gameplay in 2-year old 3D games. The other options, especially with 64 MB of memory on the 64 bit bus, look theoretically viable, but their practical value will be demonstrated only by comparative tests against their competitors.

And now the most interesting – specific frequencies and parameters:

…pay attention to quite adequate parameter sets, even if memory frequencies are a tad reduced relative to the core, (tests will demonstrate whether this was done to reduce prices or because the accelerator core couldn't digest the large data flow) and plans to manufacture mobile modules for notebooks.

And finally, the supposed position of the XG47 on the market:

Pay attention to two issues. Prices for desktop products based on XGI accelerators must be NOTICEABLY lower than those for ATI or NVIDIA of similar positioning – only in this case it'll be possible to persuade most customers, be it an end user or a large wholesaler, to take the risks and deal with a new little-known brand. But prices in the sector of inexpensive solutions depend much on the production volume, which actually places XGI into an exclusive circle position: no large volume production –> no advantageous prices –> no sales –> no large volume production. It's not so easy to break through. It would have been possible with unique features or performance, but these products demonstrate neither. Besides, ATI and NVIDIA products are already in the market and they will be eventually replaced with more advanced modifications, while XGI is just in the pre-production stage. So, this desktop legend provides no reason to be excessively optimistic.

XGI positions itself rather low in the mobile segment, on the border with graphics integrated into the chipset. In this situation the company also needs to hurry up – Intel is not asleep, on the contrary, it bites from the overall 3D pie bit by bit, even if from its bottom part. That includes the mobile sector, which is growing like on steroids due to the desktop sector reduction in many respects. Such balance redistributions are advantageous in themselves for new brands and products to break through into the market. But neither Intel nor ATI and NVIDIA were asleep and they have a lot of excellent solutions for these sectors, both ready and in the design stage. Well, we shall not be optimistic here either, but potential chances of XGI in the non-desktop sector are a tad higher – be it mobile solutions or media boxes for TV sets.

Legend Four: XG45, invisible hero of the Middle Earth

Let's approach to this question from the end:

The frequencies are good, aren't they? And the memory capacity is sufficient. Though the 128 bit limitation may be disturbing, it's actually quite adequate to the chip positioning – no matter what roadmaps tell us, this solution will actually be (if it's released) in the lower mainstream sector, especially if it comes out to the market in the same "quick and timely" way as its current development.

Now the details. Key features of XG45:

  • 130 nm.
  • DirectX 9.0c
  • Shaders 3.0 (you don't say so!)
  • 8 pixel pipelines
  • 128/64bit memory bus
  • Dedicated hardware video processor
    (deinterlacing, motion compensation, IDCT)
  • DVI (1600x1200), HDTV (1080i/720p) support
  • TV-Out and LCD panel interfaces built into the chip
  • Low power consumption (mobile solutions).

And what's the most interesting – Shaders 3.0. There was even invented a special name for this comforting fact – TrueShader 3.0 technology. It either hints that Shaders 3.0, supported in all other cards, are not true. Or it wants to assure potential buyers that Shaders 3.0 are not phony, that it's not a weak feature added for show to some pipelines (we have already come across this approach in the previous Volari generations, descending from SiS architectures).

Anyway, only tests will demonstrate to what degree these promised Shaders 3.0 will come up to real 3.0 compared to real competitors. And now we can only marvel at the audacity of the company that has taken up this difficult task. And again, long live its investors and may XGI enjoy their generosity for a long time: then, in a couple of years, we may probably see a somewhat adequate to competitors and even (again probably) revolutional product. That would be great and not so boring!

Now what concerns positioning:

It's quite up-to-date as well. The question is in price, as we have already mentioned it – how much this up-to-date performance will be. NVIDIA and ATI products are getting cheaper and more advanced, it won't be easy to lure customers to alternative products with similar prices. How much lower will it be, the price for XGI Volari? We'll learn it sooner or later I hope, as well as whether XGI drivers are stable and how many video card manufacturers will dare deal with these chips and launch serial accelerators based on them.

However, we are again pessimistic in our forecasts. Let's not be too upset and have a look at some data on XG45 performance according to XGI:

It's astonishing, this accelerator is even faster in modern shaders and offers better competition than in more conservative older applications (but if we take a closer look, it becomes clear that it would be more honest to add NVIDIA 6600 GT results to the bottom diagram as well).

We can draw a preliminary conclusion that the memory and caching system is not very productive (weak HSR and memory controller?) but pixel shader processors are rather powerful. Is it really a success, at least in this sector? Well, it's a good start, the company just needs to pull up the other aspects.

A couple of roadmaps in the end:

An attentive reader

Firstly, may note that Mainstream should be replaced with Value, and Performance should be replaced with Mainstream in these legends. At least if we want the words to represent the facts and what has been written in the first chapter of this article and shown on the first picture. Otherwise these legends will finally turn into a myth, which is not desirable ;-).

Secondly, he/she may notice that the approximate dates for the launch of XG47 and XG45 production fall at the second quarter of 2005 , XG50 – at 2006, and XG6X – at 2007 correspondingly (as we have already surmised with the most shrewd readers, by induction).

Thirdly, the XG50 will obviously improve weak points of the XG45 – there will be no special architectural innovations, just a more advanced memory controller (it will finally support GDDR2/3, it's a shame not to have this support at the beginning of 2005, even in a relatively cheap middle end segment).

The chips will be manufactured by the 90nm process technology, which will make them more attractive from the economic point of view, it will reduce heating and power consumption or increase frequencies. Note that the XG51, which will come to replace the XG47, will execute Shaders 3.0, that is will be a sterling scaled down XG50 (unlike the XG47, which functions are weaker than those of the XG45).

So, the specifications will do for the finest homes of London and Paris, sorry, Silicon Valley and Toronto. What will the prices be and will there be products? And, finally, where it is, the mythic XG45 in off-the-shelf cards with drivers, for thorough tests?

We are waiting!

Alexander Medvedev (unclesam@ixbt.com)

March 17, 2005

Write a comment below. No registration needed!

Article navigation:

blog comments powered by Disqus

  Most Popular Reviews More    RSS  

AMD Phenom II X4 955, Phenom II X4 960T, Phenom II X6 1075T, and Intel Pentium G2120, Core i3-3220, Core i5-3330 Processors

Comparing old, cheap solutions from AMD with new, budget offerings from Intel.
February 1, 2013 · Processor Roundups

Inno3D GeForce GTX 670 iChill, Inno3D GeForce GTX 660 Ti Graphics Cards

A couple of mid-range adapters with original cooling systems.
January 30, 2013 · Video cards: NVIDIA GPUs

Creative Sound Blaster X-Fi Surround 5.1

An external X-Fi solution in tests.
September 9, 2008 · Sound Cards

AMD FX-8350 Processor

The first worthwhile Piledriver CPU.
September 11, 2012 · Processors: AMD

Consumed Power, Energy Consumption: Ivy Bridge vs. Sandy Bridge

Trying out the new method.
September 18, 2012 · Processors: Intel
  Latest Reviews More    RSS  

i3DSpeed, September 2013

Retested all graphics cards with the new drivers.
Oct 18, 2013 · 3Digests

i3DSpeed, August 2013

Added new benchmarks: BioShock Infinite and Metro: Last Light.
Sep 06, 2013 · 3Digests

i3DSpeed, July 2013

Added the test results of NVIDIA GeForce GTX 760 and AMD Radeon HD 7730.
Aug 05, 2013 · 3Digests

Gainward GeForce GTX 650 Ti BOOST 2GB Golden Sample Graphics Card

An excellent hybrid of GeForce GTX 650 Ti and GeForce GTX 660.
Jun 24, 2013 · Video cards: NVIDIA GPUs

i3DSpeed, May 2013

Added the test results of NVIDIA GeForce GTX 770/780.
Jun 03, 2013 · 3Digests
  Latest News More    RSS  

Platform  ·  Video  ·  Multimedia  ·  Mobile  ·  Other  ||  About us & Privacy policy  ·  Twitter  ·  Facebook

Copyright © Byrds Research & Publishing, Ltd., 1997–2011. All rights reserved.