After the conception of our CEO Forum way back at the discussion table, we weren’t finished with what we had envisioned. Our first CEO Forum gave users and institutions that were geographically away from the Taiwan manufacturing hub the ability to understand where the market is residing and is going to reside in the near future. Granted, some of what the CEOs stated back then still has yet to be played out, but it still lent some interesting thoughts that many people did not expect.

Our original plans were to take the approach of the CEO Forum from a strategic business standpoint, so as to leave room for another project down the line. After all, CEOs are best suited to answer these types of [business] questions. The standpoint we left room was the current and future technical details relating to a company's products. However, many times CEOs are not the best people to contact for this type of stuff. (Though, they are often tech-capable.) Instead, many of the day to day duties, which involve decisions such as “which product line to pursue,” are the responsibilities for the VP of Product Marketing or someone in a similar position. However from company to company, official companies titles and responsibilities vary and get blurred. This is more a function from the way a company is structured, if anything else. And in a very simple explanation, Product Marketing is a mixture of engineering, marketing, and public relations. Product decisions, in the broadest sense, can come down to who can provide the best bid price, a product that meets specifications, and technical troubleshooting (something companies even need in the development and life of the product). Because of this, we decided to go with Product Managers as those to include in our next forum, which is to take things from a product/technical standpoint. Product Managers are engaged with consumers and are well equated with the technical side, because they deal with engineers (they often are engineers) and are involved with the product decision making process. For the purposes and reasons we cited in our first CEO Forum, we are turning to Product Managers in the HQ of each motherboard manufacturer to get a global look of all the industry trends.

The format is still the same as the CEO Forum, but we are just involving PMs (product managers) for a more technical look into industry trends. We should note that much of this took place in a month’s timeframe, about from the beginning of August. Some products were announced in this timeframe, and so some of the responses we are publishing appropriately reflect this. You will notice that Darryl Chan is "Marketing Devision Director" of Albatron, and the only one to be dubbed with "marketing" in this line up. We should make clear that he is also the head of all product managers at Albatron, but there is no official title for this position. Therefore, his official title has been listed for our purposes.

Participant - Official Title, Company.

Hunter Lee – Product Manager, ABIT Computer
Darryl Chan – Marketing Division Director, Albatron Technology Co., LTD.
Gerald Wang – Sr. Director, AOpen Inc.
Richard Liu – Product Manager, Asustek Computer Inc.
Steven Kuo – Product Manager, Chaintech Computer
Fanny Chen – Product Manager, DFI
Eric Kuo – Deputy Manager, Elitegroup Computer Systems
Calvin Yen – Product Manager, Epox Computer Co., LTD.
Chris Wang – Director of Product Marketing Division, Giga-Byte Technology
Scott Yang – Senior Product Manager, Micro-Star International Co., LTD.
Jonathan Yi – V.P. of Product Marketing, Shuttle Inc.
Tom Yang – Section Manager of Product Marketing Division, SOYO Computer Inc.
John Nguyen – Product Manager, Tyan Computer

Note: From this first run of the mill, so to speak, we have been able to test the water and have gotten some good feedback. For one, the length of the article has been consistently raised as an issue, because executives, investors, marketing personnel have found the many pages plus the amount of quotes to cumbersome and often times redundant. It is understandable that time is money for them, as they are indeed busy people. Though, for us to leave out quotes that may be repetitive in tone/thought would go against our original intent, because the range and number of quotes help emphasis the majority or the minority’s thought on the subject. So if possible, we will try and split our forum articles into two sections. This way we maintain the style and berth of input that was originally intended and keep the length of the read much lower and easier than what we first introduced. The second issue has been the topic of drawing conclusions. We have tended not to voice where the market should turn, but instead provide as much information as possible and let readers decide for themselves. Basically, letting the cards fall where they may. This is something that we will continue, all the while juggling the need for us to lend our own thoughts and insights. It is a fine line that we will continue to keep watch for, and decide on for a case by case basis. With all that taken care of, on to AnandTech’s first ever PM Forum...

1. RDRAM’s applications have died down in the mainstream market due to the lack of support. SiS is the only chipset maker with plans to support RDRAM, with its dual channel SiS658 and quad channel SiS659. Does your company currently plan on supporting...

1. SiS658 [dual channel RDRAM]?

2. SiS658 [dual channel RDRAM]?

PM #1: Although RDRAM has, in the past, been shown to be of better performance in many applications when tested against DDR technology, RDRAM support among the industry vendors has been all but eliminated, making RDRAM the technological pariah for the channel, going forward.

PM #2:

SiS659 success will depend on:

significant RDRAM performance over DDR400

RDRAM price at least 1.2 times of DDR400

availability of RDRAM to end user. If companies are able to bundle 256MB RIMM with mainboard, it will be more successful.

PM #3: Based on the market status and technology trend, we believe RDRAM might be phased out from PC market down the road. RDRAM will still exist in some niche market, such as the gaming console. But for the PC market, we think DDR memory will be the mainstream choice.

PM #4: Lack of market demands for RDRAM platforms and SiS platforms make them not suit for this segment (high end level).

PM #5: ... RDRAM seems to be dying out, even Intel gave it up.

PM #6: ... the key success factor is the total cost of ownership and the performance they get. We still think the RDRAM market is cold as ice, because the trend of memory has come to DDR/ DDRII. If, possibly, the RDRAM can go to the market, it only should be fit for the very niche market, i.e. very small quantity.

PM #7: Everyone knew that RAM bus is a good solution for high-end desktop PCs, but RDRAM has lost the chance to become the mainstream product.

Much of what is already know about the RDRAM situation is just reiterated by our responses, as 85% of product managers report no future plans for the SiS658 and 77% reported no plans for the SiS659 chipset. One of the product managers indicated that they were still watching, and therefore couldn’t give a “yes” or “no” response.

At the moment, SiS is the only significant chipset maker that has included RDRAM support in their future roadmap with their SiS658 and SiS659 chipsets, which support dual channel and quad channel respectively. Preliminary reports from manufacturers stipulate that both chipsets have memory timing issues specifically that they aren’t able to achieve the advertised memory bandwidth. However, only the proliferation of motherboards based on these two chipsets will reveal if SiS has fixed all the preliminary issues.

Since these two chipsets are supposed to be marketed towards the high-end market, they will indeed need to offer higher performance than DDR based systems. DDR is still more cost effective, and if RDRAM cannot achieve a performance lead over DDR, we will be on hand to see a very fast remission of RDRAM. As it stands, there are basically two reasons to choose RDRAM, compatibility and performance. On the compatibility side, those that incorporate the RDRAM memory interface into their designs will not have much of a choice for memory solutions, reluctantly or not. If a transition is to be made, the entire design or at least parts of it will need to be reworked, and depending on the timeframe, the switch from RDRAM to another memory solution could prove to not be cost effective. In strictly performance considerations, RDRAM can come and go easily, as preference is given to the latest and greatest. Simply put if performance is not adequate against DDR (including DDRII) solutions, RDRAM will be tossed to the curb very fast. The supply and cost issue will only accentuate the process if performance cannot be achieved. One thing is for sure, “RDRAM has lost the chance to become the mainstream product,” as on product manager put it. A less generous outlook for RDRAM was given by PM #6, who stated, “We still think the RDRAM market is cold as ice...”

We continued to follow up by asking more of a market perception question, “If yes for either, do you think SiS can be successful with RDRAM in the entry level workstation and high-end desktop market where they have just started to gain name recognition because of the SiS648 and SiS655?”

PM #1: Yes, due to price stability of RDRAM, which is an important factor for the workstation market. High performance is also a major requirement.

PM #2: No, how many end users are aware of Serverwork as a leading server chipset supplier?

PM #3: SiS is the first 3rd party Intel platform chipset manufacturer. Even though, the SiS brand been associated with the “value PC segment” for a while, we do believe the reputation of SiS products is getting better and better.

As it is, RDRAM doesn’t seem to have the good image that it is trying to reclaim. SiS is new to the entry level workstation and the high-end desktop market with their SiS648 chipset, and moving to fast with a shaky chipset may hinder their image development for this market in the future; should they fail. Though, we should note that SiS’ market perception has been gaining better momentum and credibility with the SiS648 and SiS655 and that SiS indeed has the ability to over come the “value segment” stereotype that they have been often associated with. What happens next will greatly depend on the type of products they plan to push forward with.

2. Integration has been the focus of many chipset makers, especially with the advent of NVIDIA’s nForce and nForce2 chipsets. However, one of the biggest hurdles for integration has been graphics. Can integrated graphics make the move to the entry level workstation and high-end desktop market?

PM #1: It’s still up to chipset vendors, the performance has reached a certain level, but if they can bring integrated graphics to the main stream level, there will be more of a chance.

PM #2:

For high end level MBs there is more care for
a.) Upgrade ability
b.) Flexibility
c.) Performance but not Integration matches for these key factors.

PM #3: For mainstream yes, but not for high-end desktops. At least for now…

PM #4: Even with technology changes, the market does see the difference between KLE 133 and KM400, nForce and nForce 2, 810, 845GE, and 865G. Everything is possible. One thing is for sure: it doesn’t matter where the technology goes, we will be there.

PM #5: Integrated chipsets are suitable for the entry level PC instead of the high-end PC, like PC game users, who need a powerful graphic card to execute virtual 3D effects. However, all current integrated chipsets can satisfy the task requirements of most users. I believe the integration solution will become the mainstream product due to the nano-technology in the near future.

PM #6: This integration usually means the balance between cost and performance. We believe stand-alone 3D graphic cards will still occupy the mid to high-end level graphic market. Plus the 3D graphic GPU research and development is very fast, integrated graphics usually can’t catch the technology trends due to timing and cost issues.

PM #7: No that segment requires the very high-end graphics performance, even the OpenGL performance is a must. The integrated cost/schedule, due to die-size and technology adoption consideration, can’t adopt the state-of-the-art graphics technology. This limits the possibility for the move to that segment.

PM #8: Typically, integrated graphics solutions are 2 to 3 generations behind current graphics technology, so this puts the integrated graphics solutions at a disadvantage for workstation use. In addition, the integrated solutions rarely have OpenGL driver support, which is a requirement for many workstation applications.

PM #9:

Each product segment has its own characteristics. Normally the high-end desktop PC segment focuses on performance and highly expandability. And the workstation segment focuses on stability.
The key point is integrated graphic chipsets must sufficiency support the characteristics of the workstation and high-end desktop market segments, then it can move to those product segments.

PM #10: Graphics demand for high end is ever changing and can be easily obsolete. Therefore integrated graphics is more suitable for mainstream, and entry level segment users would not mind paying a few extra dollars for on-die VGA. There is also more volume in the low end segment, which gives chipset makers faster return on investment.

PM #11: I think entry-level workstations don’t need high-end graphics ability, but high-end display processing does.

It seems that the majority of product managers (69%) believe that integrated graphics isn’t going to make the cross to entry level workstations and high-end desktops any time soon. The only non “yes” or “no” response came due to the comment, “Yes for entry level WS, No for high-end desktops.”

This isn’t to say that it won’t ever happen, but it won’t be happening soon. As PM #8 stated, “Typically, integrated graphics solutions are 2 to 3 generations behind current graphics technology, so this puts integrated graphics solutions at a disadvantage for workstation use.” This plus the requirements for OpenGL support will make it hard for the entry level workstation and high-end desktop market to accept integrated graphics.

Performance is one thing, and the ability to upgrade is another, as integrated graphics does’t have the same upgrade characteristics as stand-alone graphic cards. This ability to upgrade also translates to cost, because even if integrated graphics reach the computing capacity of the R350 core, upgrades will basically require the overhaul of the motherboard. This bodes very well for ATI and NVIDIA, as the need for stand-alone graphic designs will continue to exist for the foreseeable future from a cost and performance standpoint. However, IGP solutions from ATI, Intel, and NVIDIA will still need to continue to improve, as the base market for IGP solutions and the minimal graphic computing requirements continue to rise.

3. Lately, AMD has been emphasizing that the industry is ready for 64-bit computing, in reference to their upcoming Athlon64. Do you think the Athlon64 has enough 32-bit benefits to be immediately successful?

PM #1: Yes, we believe AMD might have a great chance to succeed with the Athlon 64 CPU. But it all depends on the quantity of CPU they can supply for the market. If the quantity of CPUs is low and limited, AMD may lose their chance to get back market share from Intel.

PM #2: Customers are going to question the practicality and benefits of purchasing a 64-bit processor with a higher price tag over the 32-bit only counterpart, Athlon XP. The main driving reasons behind successful adoption of the new architecture among consumers will be price, performance, and compatible applications. For example, if Microsoft and other software vendors cannot provide compelling applications (i.e. applications that take advantage of the new architecture to provide significant performance enhancement, and applications that have backward compatibility with 32-bit counterparts), there will be no reason for consumers to upgrade.

PM #3: The key should be the readiness of 64-bit applications. AMD will launch the new CPU with future features, but the better way to promote the CPU is to exploit the performance that user can be benefit from the new stuff.

PM #4: Need OS and AP support. Need more time for ramp up.

PM #5: Prior to the end of this year, the available quantity of AMD64 is not so much, so we don’t expect the selling quantity to be that big, but 64-bit computing is a good selling point to compete with Intel, if the price of AMD64 is aggressive enough, it will be successful in a short time.

PM #6: Now the CPU yield rate will be the major problem.

PM #7: The success of Athlon 64 will depend on compatibility with DDR memory. Its performance is not the major concern.

PM #8: “The next big thing” for the PC industry in the rest of 2003 is AMD 64-bit processor. We believe it will benefit the world-wide PC economy. It will take off and it must take off.

PM #9: Not till we see a 64-bit OS released. Before that, we will probably never see the “true” performance of it.

PM #10: Yes, people are not just looking at current performance benefit brought by Athlon64, but also it’s headroom for future upgrade, I mean its potential for higher frequency and performance is much better than K7, or even P4.

PM #11: But the CPU supply might be a concern.

The feedback is a bit split both ways, but a little over half of the product managers, 54%, give AMD the benefit of the doubt. Only 15% don’t give a solid “yes” or “no” answer, but instead cite concern for the availability of 64-bit applications and compatibility of Athlon64 with DDR memory. Regardless of their opinions, the majority still points to two obstacles that Athlon64 will need to overcome: software and supply. On the software side, there will definitely need to be more support for Athlon64 to really ramp up AMD sales, either near product launch or soon after. And as PM #1 writes, “If the quantity of CPUs is low and limited, AMD may lose their chance to get back market share from Intel.” So far, AMD has done a good job of marketing that 64-bit processing is the next step for the industry, which is translating into demand. However if AMD can’t actually supply the product on schedule, to a degree everything else is going to be irrelevant. Many of these product managers see this as a rare occasion for AMD to gain back mainstream market share from Intel, because it isn’t often that such a generation jump takes place nor has the market been in this specific situation before.

Regardless of these opinions, something is going to happen soon. Athlon64 is going to do something for AMD. At this point, it can’t really fail, only fall short of expectations. But many are putting it the way PM #8 did, “‘The next big thing’ for the PC industry in the rest of 2003 is AMD 64-bit processor. We believe it will benefit the world-wide PC economy. It will take off and it must take off.”

4. Upgrade cycles are stretching out and i845PE motherboards have enjoyed a high volume of sales. For the next two quarters, is your company…

1. planning to do smaller volumes of production for 865G/865P/865PE motherboards?

2. expecting low sales for 865G/865P/865PE motherboards in the next two quarters?

PM #1: Because of the economy, users/corporations still have a tight budget. However, since new Intel FSB800 CPUs have become mainstream, the demand of new 875/865 family boards also becomes strong and stronger. …we will definitely extend our production to fulfill the demand.

PM #2:

1. The price gap getting less in clone market between FSB 800 vs. 533 CPU.
2. The price gap getting less between Serial ATA HDD vs. ATA/133 HDD.
3. 848P platform lower price will squeeze 845 series portion.

PM #3: 848P will share some market of the 865PE.

PM #4: Intel will promote the 865 series more aggressive in 2H 2003, including price cut downs.

PM #5: 845 series demand will stay strong in 3rd world countries, basically, but the 865 series will be the mainstream choice for industrialized countries.

PM #6: To see an increase of volume for 865G/865P/865PE, we need to have its chipset cost approach the $25 mark. This will make the Springdale chipset more mainstream.

PM #7: We except the market will start moving to dual channel DDR instead of single channel. Dual channel will become mainstream in the following quarters.

PM #8: It all depends on how Intel is gonna push its 800MHz CPU.

PM #9: The new chipset i848P will be a sweet spot from Intel. The transition planning from 845PE to 848P is POR to happen in Q4. The 845PE will EOL gradually.

The response we got back for both parts of the question were the same: 85% yes and 15% no. We asked this question to see if the high sales of the i845PE/GE motherboards have saturated the market to the point where upgrades or sales of the next generation of chipsets would be hindered. It seems that isn’t the case. Manufacturers are, for the most part, gearing themselves for relatively the same volume of production. Additionally, we are on hand to expect more price cuts in the last half of 2003, which bodes very well for consumers and manufacturers. It was once suggested that dual channel would be skipped for DDRII, but this hasn’t materialized. DDRII is still a ways off, and dual channel support is up due to new Intel chipsets. It is suggested from the feedback that single channel DDR will hit the end of its life sometime in Q4, which means sales will be further boosted by this phase out.

5. Opteron will have applications in both the server and the high-end desktop market due to its low cost. Do you expect the high-end mainstream market to be receptive of it?

PM #1: But it depends on its yield rate.

PM #2: Yes, as long as the operating system and server manufactories can prove how reliable and stable the Opteron system can be. We believe AMD will step into this niche market successfully.

PM #3: The high-end mainstream market will be receptive of the AMD 64 (939 pins).

PM #4: The latest information from AMD shows that there will be 939-pin Athlon64 with dual channel support in 2004, with new ball out the 939-pin is optimized for 4-layer M/B, which may have a better chance for the market.

PM #5: People expect good technology to enhance their productivity. Also, high current/projected value is a key to the success.

PM #6: It’s expected that Opteron will be embraced by the high-end mainstream market, especially since support for the platforms will be coming from vendors such as NVIDIA, who have the expertise and branding to provide muscle behind educating customers on the benefits and applications of the AMD64 architecture.

PM #7: The cost of Opteron is around 400-500. This factor is due to the cost register-DIMM. This will hit the niche market, but will not go to the mainstream.

PM #8: Open architecture is must and more popular OS can share common resource.

It is hard to tell how the Opteron will fare in the high-end mainstream market. Certainly, the price point is and will remain to be one of the big issues, and most of the product managers seem to be indicating that Opteron will not cross over in this respect. Instead, Athlon64 will be the choice for this borderline market. AMD hasn’t really clearly defined which product will be marketed towards the high-end portion of the mainstream market. It has been suggested that game enthusiasts and those types of people that wanted to use RDRAM but were looking to an AMD product will choose Opteron. What is more likely is that the high-end market will pick up Athlon64 when cost is an issue, and the smaller segment of that which is concerned with performance will turn to Opteron. Cost is a relative issue, as Opteron is cheaper than Intel server processor solutions, but it still is going to have a good cost premium on Athlon64. For this reason, reception for Opteron in the high-end mainstream market will likely remain low but much of this still will depend on how Athlon64 pans out. From a cost perspective, Opteron will be better suited for four and eight way servers, even though eight way servers make up less than 5 percent of the market. Dell has already cut their support for this segment because it is too small. Opteron will have a clearer advantage at these segments because the cost of multiple processors will quickly add up, as well as due to the reduction of competition.

On a side note, Intel has recently announced their new budget Itanium processors, officially dubbed low voltage Itanium 2 (1GHz with 1.5MB L3) and Itanium 2 1.4GHz (1.5MB L3). The low voltage and 1.4GHz price point [in the mid $700 US and lower $1000 US range, respectively] is Intel’s response to help close the price gap that Opteron created an issue of. The high-end Opteron is also in this range, so there will be higher competition compared to the earlier Itanium 2 processors that were just shy of $5,000 US.

Final Words

We would again like to thank the participants and their respective companies for helping make our PM Forum a reality. Your time, effort, and cooperation are greatly appreciated. Our dream to help create a better industry will continue to be a reality, so long as we work together.