WindowsNT Intranet Solutions Conference

Andrew S. Grove
San Francisco, CA, USA
August 27, 1996

Good morning. Today I'm actually going to give two speeches in one. My first role, if I could put my first hat on, is that of the person who is responsible for Intel's own IT organization. In that capacity I would like to describe to you a little bit of Intel's real life case history of what we have recently done that is particularly relevant to this conference. First let me characterize Intel as an IT organization. We are a Fortune 60 something company and correspondingly we have a fairly large IT organization and a good-sized IT budget, something in the vicinity of $700 million. Essentially all people with a desk have a desktop computer. Quite a number of the field team also have a mobile computer, but the desktop clients alone are in the range of 50,000 units.

Approximately nine months ago, we decided to go Windows*NT across the entire organization, servers and clients both. I would like to give you a little bit of a background leading up to that decision because I think it is pertinent and there are possibly observations in there that might be applicable to some of you. First of all, one thing that has allowed us to make the decision is that our current desktop configurations are fairly muscular. By the end of this year we are going to be practically 100% based on Pentium® processors, practically 100% 60 megabyte memory or greater, and a half a gig disk or greater. From a hardware standpoint, these machines, do not require an enormous amount of upgrading to accept Windows NT.

As an organization, we have increasingly used our client/server network for significant and increasingly mission critical applications. Consequently when we really surveyed what is truly important for us in the use of these network computers in our own businesses, the three criteria came up in this order; performance, reliability and security. In each of these cases Windows NT provided a step in the right direction for us. Our own measurements on the type of applications that we use, indicate that when we use Windows NT and 32-bit applications on a comparably equipped client, we got a 15% to 20% improvement in performance versus Windows*95 and 16-bit applications. Regarding reliability, the most important step is consolidating to a single network operating system. During this transition throughout the enterprise, the various network operating systems are still able to operate.

Regarding security, the most important development was having the same operating system on both the client and the server, allowing us access to security services which previously weren't available.

The most surprising element of the study that has resulted in this decision was that this choice, was an economically favorable one as well. We have undertaken a fairly serious study of the total cost of ownership per desktop computer across our enterprise and the result was that our Windows 3.1 personal computer costs ranged $9,000 per year including end-user support costs. Had we made the transition to Win 95, we would have gained a little less than $1,000 per year improvement. With the movement of NT, our anticipation is that we are going to gain almost $3,000 per year improvement due to the factors that I described above. The cost penalty in terms of hardware, even for new clients as you see the client base continue to grow, is negligible in comparison to these numbers.

What we have found is that by deciding on a completely NT workstation and NT server across our corporation, we have improved manageability and reliability; improved performance for 32 bit applications; simpler upgrading of the machines with a lot of down the wire upgradability possibilities available to us. Consequently a combination of these as well as increased capabilities, will lead to increased possibilities of down the wire management, and a reduction in support costs by 30% to 50% by the end of this conversion, which is going to take us two years. So these are the supporting arguments for our decision which is now being implemented with a portion of the population already having been converted. The first site, that we completely converted to Windows NT based environments, spontaneously gave an Intel quality award, (which is something that we do internally), to the IT organization for the exceptional results that they achieved with this conversion. It is not very often that an IT organization gets a quality award from their users!

So now I would like to switch my hat and take the standpoint of a supplier of infrastructure products or building blocks that are pertinent to moves like this and the deployment of Windows NT. There are two subjects that I would like to address. The first is to talk a bit about our platform road map and secondly about some of the major trends in business computing as I see them and relate them to our strategy. The evolution in business computing is very evident so far. It has gone from the basic stand alone PC used for the well known stand alone productivity applications, such as word processors, spreadsheets and the like, to today's business personal computer. This is typically a Pentium processor running Windows 3.1 or Windows 95, graphical user interfaces, network, with typically e-mail and file transfer being significant applications. We are beginning to see a transition of the current business computer to something that I will simply refer to as the "new business computer". This "new business computer" will increasingly be using high performance processors exemplified in the Pentium Pro processor. It is going to have higher bandwidth capabilities, such as 100 megabit Ethernet; very importantly, it will feature built-in manageability, and as a major part of its use, it is going to have Intranet access capabilities and increasingly will be able to demonstrate the utility in business application of advanced 3D graphics.

(At this point, Dr. Grove does a demonstration running ordinary business applications, on a current business PC compared to a new business desktop configured for this purpose. For this demonstration, we have two systems which are very similarly configured except for a fundamental difference. One is a Pentium processor running at 133 MHz; the other is a Pentium Pro processor running at 200 MHz. Both are using Windows NT workstation software as their operating system as well as a few of the commonly used business desktop applications such as PowerPoint, Excel, an optical character recognition software, a fractor limit viewer, a desktop database, etc. The demo initiates an automated script, which will run through these applications, unravel some data, and show how much faster a Pentium Pro processor based system handles the applications vs the Pentium processor-based system. The sequence of events in the demo are first an OCR on a scanned image, which is embedded in a document in the database. At that point, both the systems will be running one task only. Following that, the systems will be loaded with a multi-tasking load, which includes things like simultaneous wire scanning and printing. The Pentium Pro processor-based system finishes the tasks in 1.3 minutes; whereas the Pentium processor based system took 2.3 minutes.)

The Pentium Pro processor-based machine represents today's business computing machine, and I have a few data points that support this statement. We have in fact ramped the Pentium Pro processor into production faster than any other previous processor. Equally significantly, the personal computer industry ramped it into the volume mainstream faster than any other processor based machines before. To illustrate this, it took 50 months, before the 486 based machines reached the $2,500 fully equipped price point -- which I arbitrarily picked as the mainstream machine cost. With the Pentium processor we reached this price point in approximately five quarters. But the industry did a lot better with the Pentium Pro processor. Not quite three quarters after it's introduction (in November 1995) you are seeing $2,500 machines on the market. So it is a mainstream machine, it is affordable, and we're going to do our best to bring it down on the cost curve and enable its volume in the quarters to come.

Of course the Pentium Pro processor has been accepted as the fundamental microprocessor for servers right from the beginning. Dual and four way symmetrical multi-processor servers have increasingly been appearing on the marketplace. To build these servers, we have integrated a variety of additional technologies, these include a special server PCI chip set, as well as the intelligent I/O chips, (abbreviated I2O), that deliver this performance through the I/O subsystem to server applications as well. We have dedicated ourselves to bringing the same kind of cost effectiveness that we have achieved with desktop personal computers to the server environment through the introduction of standard building blocks. Here too, they will consist of the microprocessor, the chip sets, the motherboards, and in many instances pre-packaged servers that we supply to a variety of computer manufacturers. We are in fact very pleased that the standard high volume server configuration has become a broadly accepted item throughout the industry, supported by companies like Digital, Tandem, Fujitsu, Unisys, and many others.

We are also at the point in time where we have a major event looming ahead of us on the processor road map. This has to do with the introduction of our MMX˜ technology. MMX technology involves adding a set of instructions to the Pentium processor, which particularly accelerate the performance of multimedia applications and communications, specifically Internet related multimedia applications. It is the first major modification of the Intel architecture instruction sets since the 386 processor. It involves adding 57 general purpose instructions and high-clocked operations. The benefits of the MMX technology is that a variety of applications such as conferencing, image processing, video, audio, modem applications, speech recognition, and the like, achieve an average of 20% to 50% improvement when the applications are optimized for MMX technology. This corresponds typically to the kind of improvement that you get due to architectural design factors between two generation of microprocessors.

This kind of application acceleration performance has actually manifested itself in another fashion. By offering this type of performance built into every platform, we can achieve substantial simplification and cost reduction at the platform level. Today most of these functions are performed through the addition of accelerator boards, which brings both cost to the hardware platform and complicates the support costs that we were dealing with earlier. These costs go up in more than a linear fashion, considering the number of add-in cards that we have and subsystems that we have in the system.

We will be introducing the MMX technology fairly rapidly to the mainstream as well. Just to illustrate that this is a chart that shows our growth rate of microprocessor production. The personal computer market in '95 was estimated to be approximately 60 million units. Our plan is to convert the entirety of this microprocessor stream to microprocessors equipped with this MMX technology instruction extension starting early '97 and ramping all the way through '97 and '98, completing the ramp by the end of '98. This is a significant commitment on our part because these additional instructions require extra silicon surface area, resulting in a larger die size. This corresponds to having to deploy one extra factory in order to introduce this instruction set across the entire line. One factory for Intel represents an investment of approx $1.5 to 2 billion. This is a very, very large commitment. We undertook this commitment of going across the board with a more powerfully accelerated microprocessor range because we are convinced that multimedia elements in both stand alone and communications applications will be so crucial that this acceleration will be necessary.

In summary, we see the improvement in performance introduced by the deployment of MMX technology in the Pentium processor range. We will combine this improvement with the addition of the dynamic execution capabilities of the architecture that's deployed in the Pentium Pro. Later, we're adding another step in performance on the Pentium processor with MMX technology to bring that additional step into the mainstream marketplace starting next year as well. So that's what's coming from Intel in the platform area.

I would like to spend a few minutes talking about some of the trends that I see currently driving business applications and business computing. I also want to touch on the growth of medium sized businesses, Intranets in big business, manageability and visualization. When we look at the worldwide computer market, the personal computer market, and try to segment them by home use, medium and small business, and large business, and we use a completely arbitrary definition of 100 computers equals 100 employees. Based on that, the market segments in the following fashion. The yellow represents the home user, the red represents the medium to small business, 99 employees or below, and below that you have the enterprise and the government. You see in '96, according to these estimates, actually the small and medium sized businesses may be consuming a larger number of personal computers than the larger enterprise organizations. This is supported by U.S. data coming from IDC worldwide research. It suggests that the '96 growth rate of the business segment - actually the small and medium business segment, is in fact going to be the highest of the three business segments. Considering this rapid growth and the substantial size of this market, along with the observations that this market tends to purchase performance with head room, this is in my opinion, an excellent target area for the industry to introduce new business desktops. Those of you in the audience who make your living developing Windows NT applications should consider targeting this business segment for custom Windows NT development. I wanted to point out the emergence of the small and medium sized businesses as a significant market to target.

My second point is the appearance of Intranet in big business. This is very evident to me from looking at what is going on in our organization, which is probably not all together atypical. We see a variety of stove pipes of environments that are different between functional specialties, like order entry, manufacturing and materials management to name a few. We see a multiplicity of environments with a multiplicity of databases. In many instances there are completely different human interfaces and a very contorted transition of data from one environment to the other. Each of these applications was developed at different points in time in our history. Each of them were developed with different technology, and consequently their integration is not impossible but it's extremely difficult.

The introduction of the web browser, the concept of a browser as well as Internet technology, allows us to scan across these organizations, these "stove pipes", and provide us with a uniform user interface with which an end user -- without additional training - can access data residing in the different databases and the different structures easily navigating them using relatively simple, low cost training methods. And given that I characterize our environment where we have in fact reached approximately a parity between the number of employees and the number of desktop applications, the number of people to whom this simplicity is relevant equals the total number of our employees, which is a very large number.

The consistent desktop interface, for access to and distribution of data to and from heterogeneous databases, drives the application and the deployment of the Intranet at Intel and many other interfaces. The story doesn't end there because with this deployment, we will increasingly the application of object oriented technologies, which render each client at various occasions even the server itself. As we continue to move towards an object oriented environment, the distinction between client and server will diminish, necessitating or conversely taking advantage of the computing power that is already resident and will be resident on the desktops of the end users. So this combination, the suite of Internet applications and the near term deployment of objects is completely consistent with what we have been talking about.

The next trend that we see is manageability, without which all of the cost figures and cost simplifications that I indicated earlier in the presentation that we hope to take advantage of would not be possible. And network management, which is present in all mainframe centric, classical computer applications, is rapidly coming into the distributed environment based on networked PCs.

I have some numbers that I would like to share with you.

The number of networked PCs deployed in organizations today is in the range of 120 million. The portion of these PCs that is manageable today is in the 10%, 15% range but that kind of 15% is brand new. Two years ago it was practically zero. The conversion of networked PCs to networked and managed PCs, -- fully instrumented PCs -- is one of the fastest trends in business computing today. A key solution to achieving that for well managed and manageable desktops is the introduction of an industry standard instrumentation, which after a number of years of work by the desktop management task force (DMTF), is a probability at this point. I can't quite say certainty but it's well on its way to that in the form of the desktop management interface, or DMI. DMI, Desktop Management Interface which is the instrumentation agent that is present in the systems and subsystems, which is the agent that is embedded in group management software like Intel's LANDesk™ or system management software, like IBM's Tivoli. Through this agent, one can actually run a well managed network, which is the key to achieving the cost of ownership benefits that I indicated earlier.

Visualization is another major trend in our business computing environment. It is driven by the complexity of the networks, the complexity of the applications, and the complexity of the desktops. And visualization is also driven by something very simple, the fact that we all are able to digest information much more rapidly in a visual form. We are able to retain a visual image of a complex phenomenon, like a network, much more rapidly. Of course the fact that the Intranet introduces the browser metaphor, which is visual in itself, sets the stage for further deployment of visualization.

(At this point Dr. Grove demonstrates the concept of visualization in an IT by using a Pentium Pro processor-based 200 MHz Intergraph workstation running Computer Associates Unicenter TNG. This represents a robust 3D configuration in a today's business computing environment.

The demo shows how, using the Internet, a systems operator can remotely detect a failure in a server system, through the use of objects, rather than text.

This demonstration described a series of IT issues; situations which happen in an IT environment, such as the door of a system being open, the plug being removed, a more serious concern such as overheating or a fire within the system at a remote location.)

This 3D environment accurately represents the enterprise or Intranet, by attaching texture and images to these objects that we're going to manage.

Visualization is in fact so important that we are devoting a considerable amount of effort to build it into our microprocessor and chip set plans. We set ourselves a target of introducing a 50% improvement in visualization and to basically reach what is considered an arcade quality video. We have a ways to go from what you have seen here, into our desktop platforms in the relatively near future.

In closing I would like to leave you with a couple of requests or suggestions. If you are a Win NT developer, I would like to suggest that you build your applications taking advantage of the platform development that I described here; that your target platform should be the "new business desktop computer" based on the Pentium Pro processor, because we intend to make that the common business desktop of choice.

I would like to call your attention to MMX technology, already on the horizon, bringing benefits to visualization, multimedia and communications applications. I would like to suggest that you take advantage of the manageability that's provided by Windows NT and a combination of Windows NT and software that is built upon a desktop management interface, a DMI. And for those of you in the audience who run IT shops like ours, I hope I made some progress in demonstrating to you that the combination of Pentium Pro processors and Windows NT delivers maximum performance for business computing, which is very appropriate and timely given today's business megatrends that we are all dealing with. Thank you very much.

* Other names and brands may be claimed as the property of others.