Recently at MaximumCPU, we were able to have a nice chat with Dan Snyder, a representative in the Technical Public Relations department at the Intel Corporation. It was a great chat and we’d like to thank him for making it happen once again. Most of the questions were things that I, Steve, had planned out, but Nick was also a good part of the interview. He can verify that things were said aren’t being made up. The interview was a great insight into why Intel does some things it does and about how its planned for the future.

At Mr. Snyder’s request, we will not be posting a word for word account of the interview, but key parts instead. I will supplement what he had to say with things already widely available to the public, and hopefully we’ll see what’s in store for the future of computing on many levels as well as part of the business aspect of Intel. Overall, not all things said in this article will be Dan’s words. Some of it will be personal opinion and speculation made by the editors. Wherever Dan said something, it will say that Dan Snyder said it. Please don’t write nasty e-mails to Mr. Snyder because you think this is all him, all the nasty stuff can be sent to me instead.

Intel is, as anyone reading this article probably knows, a multi-national company that makes millions of dollars a year in revenue and spends a great deal of that money on research and development (R & D) of their products. One thing in R & D is flash memory. When I asked about Intel’s interests in the flash memory market, and the recent onslaught of press coverage about it, Dan didn’t tell us much. That’s okay though, because Dan’s a CPU guy and in a direct quote, Mr. Snyder said, “CPU’s are our bread and butter [at Intel].” . However, he did say that there are a lot of people working on several memory technologies in Intel’s labs. Intel would want a place of flexibility in that market, like it established with its three lines of CPUs, Dan commented later on in the interview. Naturally, we assume Dan meant four lines of CPUs, being: Itanium, Xeon, Pentium, and Celeron. It is possible he also meant CPUs for servers, workstations, and desktops. (Do laptops/notebooks fit in too?)

Dan would not comment on the fact that AMD has created the standard of using NOR flash memory. I was not surprised, since no company comments directly on their competitors. But seeing as they are researching all the possible flash memory options and have a “ton” of guys working on it, maybe Intel has a new solution up their sleeve. Only time will tell us.

From what I’ve seen on a personal level though is that CPUs are the rage, and the new buzz around computer enthusiasts has been 64-bit computing. Right now, on the Intel front, there is only the Itanium. The Itanium, as Dan Snyder pointed out, is catching on in the server and enterprise market very well. It’s about time really, since the technology has been a while coming to its potential. Many top-level distributors are interested in marketing this chip in their server lines if they aren’t one of the many who already are. The cost, however, hinders this CPU from being a desktop component at the time. For enterprise solutions that require large bandwidth and a lot of data to be moved, the new Itanium 2 line is designed for them. OEM computer distributors will have the Itanium to add to their lines of high-end servers and workstations, and numerous software titles are ready for renewed interest in the Itanium. Dan seemed very optimistic that the Itanium is going to change from what people consider a bust to a great computing platform.

What about desktop 64-bit CPUs though? Over at the site Chip-Architect, and especially on this page, it seems the Prescott core that is currently part of the Pentium 4 line is very easy to convert to a x86-64 CPU part. The capability is already there, Dan went on, to convert the Prescott from 32-bits to 64-bits with relative ease. This proves the article written in 2003 from Chip Architect. That website, however, refers to the CPU core in question, the Prescott, as the Pentium 5. Dan Snyder said that the 64-bit Prescotts will still be called Pentium 4’s, but with a three digit name instead of a MHz or GHz number at then end.

These new CPUs will be clearly marked. Instead of having the Pentium 4 3.0 GHz, you’ll have the Pentium 4 ***, with the numbers (currently stars) representing a set of features collectively such as the clock speed, front side bus speed and amount of L2 cache. With this new naming scheme, there will be no confusion about what chip you are getting. While current Pentium 4’s spread across two sockets, 3 front side bus speeds, and three cores, the new naming scheme will make it clear what CPU you are using. The first chip using this new scheme was recently announced in Intel’s mobile lineup. In fact, he even said that naming chips was something that happens relatively late in the development process. This surprised me, but was not unbelievable.

These 64-bit Prescotts will need new motherboards. As well as motherboards, they will be in a new package. The LGA775, standing for land grid array and it has 775 pins. You’ve most likely seen the new motherboards where the pins are on the motherboard themselves. The chips have 775 contacts on them, which fit onto the pins of the motherboard. This will lower the amount of returned CPUs Intel has, but raise motherboard returns for bent pins. However, Dan told us that the die will have little to no changes at all to fit within the new package. Hopefully this will keep cost down for OEMs, distributors, and the end-user of course.

While the die of the Prescott remains the same, the connectors go up from 478 to 775. Extra connectors are assumed to be need for 64-bits, but grounding is also a concern for sure. Dan stressed the importance of grounding during the interview very heavily. He hinted upon the fact that more grounding pins mean more headroom and system stability for the long run.

As it lies right now, it looks like a desktop 64-bit option from Intel is not far from happening. Xeon’s will have the same option as well, as the workstation is where the technology is truly needed. Another thing Dan commented very heavily upon was that right there. The workstation has the true need for 64-bits, as well as multi-cores and multi-threads that everyone seems to love. The truth is that Intel is shooting to have multi-core CPUs for 2005, but I couldn’t get a better timeframe then that. Dan did say that the multi-core CPUs will not be based on the Prescott die, as various other sites have been reporting. At least I got some answers for myself, even if I’m not the first reporting it.

When I asked him why the delay was taken and why the Prescott was not released as a 64-bit part right away, his response was very simple. Drivers and operating systems were not ready. Sounds simple enough to me, and I would like to thank Dan for being so blatant. Simply put, without drivers and an operating system, there is no need for a 64-bit CPU. On top of that, there aren’t many applications that can support 64-bits. Basically, it would be a waste for Intel to produce and for the end-user to buy. It’s better for current users to stick with Hyper-threading.

Most people into computers know that hyper-threading allows for a virtual dual processor. It runs two threads at the same time, so while you get a performance increase as if you had dual CPUs, it shares resources such as L1 and L2 cache in addition to system memory. This means that you don’t get much benefit as you would with a true dual-CPU system. Dan said anywhere from 5 to 20 percent is the average increase from using Hyper-threading, but mostly on the lower edge of those numbers.

The applications that are multi-thread and multi-CPU capable are not truly ready though. Why would Intel use Hyper-threading in an ad campaign when people can’t use it? Well, said Dan, certain applications are already ready. Photoshop and 3D Studio Max were examples he used, saying they are ready for the future today. There are several elements, as Dan called them, come into play with Hyper-threading. The first thing is how most applications use one thread at a time. However, with the proper versions of Windows, threads are better managed. This allows one thread to be one application while the other thread is running a second application. Multitasking is a lot smoother as a result of this. Intel, as always, according to our representative, is concerned with the everyday needs of someone using a computer.

Multitasking is when you are doing more then one thing at once. Dan used an analogy of using an instant message program, listening to an MP3, and writing a word document at the same time. This type of instance is where the most people would see a difference. It’s about real world performance with Hyper-threading and not about the small gains in benchmarking programs. This is the reason for the ad campaigns involving Hyper-threading. People really do use it.

Multitasking also seems to be at the center of the reason for multi-core CPUs. Intel plans on releasing the first dual core CPUs, as I already stated, sometime in the next year or so. This is also the time of the 65-nanometer process to be released. Dan wouldn’t say whether they will be in the same product, but one can only assume such a thing. The dual core CPUs will have some major advantages over current CPUs. Multitasking is already one we know about. The separate resources are also a vast improvement over Hyper-threading, where resources are shared. According to Dan, separate dedicated L1 and L2 cache and system memory (forced dual channel?) will speed a PC up greatly.

While Intel is in the midst of all this dual core business, they have no plans to integrate a memory controller. Other companies have had great success in doing this, most notably AMD (a company Intel won’t comment on), but Dan reminded me of the eventual lack of market acceptance of Rambus. Will this cause a performance decrease compared to other CPUs on the behalf of Intel? But Dan does make a very valid point. Instead of changing the entire CPU for a new memory controller, the northbridge will allow for a change, as in the Rambus to DDR situation not all that long ago. Perhaps the downfalls are greater then the advantages in the eyes of the engineers.

These dual core CPUs will be released in an unknown socket, but the form factor is not in question. BTX is the way of the future for Intel, as well as a large group of other people. The spec sheet is publicly available to anyone at this website in a PDF file. The LGA775 socket will also be released in BTX form factor and it seems that cases and power supplies will be readied, or at least that was the impression Dan gave me. The BTX form factor changes the locations of many connectors that we have come to grow and love as a community. Well, maybe not love, but we all know the basic layout of a motherboard. The BTX standard is more balanced, as per its acronym. The way heat is dealt with his different. This is a very good thing, in my opinion at least.

The socket 478 Prescott is already heating up computers much more then former Pentium models ever have. With the dual core CPUs coming out as well, the thermal output of the CPU would have to go up. There is no logical way it won’t raise. BTX will deal with this heat much more efficiently. On this topic, Dan said that he power requirements and thermal envelope will not go down at all for current Prescott core CPUs, even with the socket change. The way heat is dealt with will help make platforms with the new socket cooler however.

For future CPUs, Dan told us that Intel’s desktop people are looking more at the mobile chips and how to take some tips and tricks for their own CPUs. Other websites already have reported that the first dual core CPUs will be the mobile CPU with two dies on one chip. Dan would not comment on that statement at all, so it may or may not be true. Power consumption is lower, as is the thermal envelope on mobile processors, so it’s a good start for Intel to lower these two things.
 

All in all, the future looks bright for computers and Intel as a whole. If there’s anything that any of our readers would like to hear about, drop an e-mail to either Nick or my self and we’ll see if we can get any answers. Once again, thanks to Dan Snyder for taking his time to talk with us. Until next time, I’m Steve for MaximumCPU.net.
 

When I asked him why the delay was taken and why the Prescott was not released as a 64-bit part right away, his response was very simple. Drivers and operating systems were not ready. Sounds simple enough to me, and I would like to thank Dan for being so blatant. Simply put, without drivers and an operating system, there is no need for a 64-bit CPU. On top of that, there aren’t many applications that can support 64-bits. Basically, it would be a waste for Intel to produce and for the end-user to buy. It’s better for current users to stick with Hyper-threading.

Most people into computers know that hyper-threading allows for a virtual dual processor. It runs two threads at the same time, so while you get a performance increase as if you had dual CPUs, it shares resources such as L1 and L2 cache in addition to system memory. This means that you don’t get much benefit as you would with a true dual-CPU system. Dan said anywhere from 5 to 20 percent is the average increase from using Hyper-threading, but mostly on the lower edge of those numbers.

The applications that are multi-thread and multi-CPU capable are not truly ready though. Why would Intel use Hyper-threading in an ad campaign when people can’t use it? Well, said Dan, certain applications are already ready. Photoshop and 3D Studio Max were examples he used, saying they are ready for the future today. There are several elements, as Dan called them, come into play with Hyper-threading. The first thing is how most applications use one thread at a time. However, with the proper versions of Windows, threads are better managed. This allows one thread to be one application while the other thread is running a second application. Multitasking is a lot smoother as a result of this. Intel, as always, according to our representative, is concerned with the everyday needs of someone using a computer.

Multitasking is when you are doing more then one thing at once. Dan used an analogy of using an instant message program, listening to an MP3, and writing a word document at the same time. This type of instance is where the most people would see a difference. It’s about real world performance with Hyper-threading and not about the small gains in benchmarking programs. This is the reason for the ad campaigns involving Hyper-threading. People really do use it.

Multitasking also seems to be at the center of the reason for multi-core CPUs. Intel plans on releasing the first dual core CPUs, as I already stated, sometime in the next year or so. This is also the time of the 65-nanometer process to be released. Dan wouldn’t say whether they will be in the same product, but one can only assume such a thing. The dual core CPUs will have some major advantages over current CPUs. Multitasking is already one we know about. The separate resources are also a vast improvement over Hyper-threading, where resources are shared. According to Dan, separate dedicated L1 and L2 cache and system memory (forced dual channel?) will speed a PC up greatly.

While Intel is in the midst of all this dual core business, they have no plans to integrate a memory controller. Other companies have had great success in doing this, most notably AMD (a company Intel won’t comment on), but Dan reminded me of the eventual lack of market acceptance of Rambus. Will this cause a performance decrease compared to other CPUs on the behalf of Intel? But Dan does make a very valid point. Instead of changing the entire CPU for a new memory controller, the northbridge will allow for a change, as in the Rambus to DDR situation not all that long ago. Perhaps the downfalls are greater then the advantages in the eyes of the engineers.

These dual core CPUs will be released in an unknown socket, but the form factor is not in question. BTX is the way of the future for Intel, as well as a large group of other people. The spec sheet is publicly available to anyone at this website in a PDF file. The LGA775 socket will also be released in BTX form factor and it seems that cases and power supplies will be readied, or at least that was the impression Dan gave me. The BTX form factor changes the locations of many connectors that we have come to grow and love as a community. Well, maybe not love, but we all know the basic layout of a motherboard. The BTX standard is more balanced, as per its acronym. The way heat is dealt with his different. This is a very good thing, in my opinion at least.

The socket 478 Prescott is already heating up computers much more then former Pentium models ever have. With the dual core CPUs coming out as well, the thermal output of the CPU would have to go up. There is no logical way it won’t raise. BTX will deal with this heat much more efficiently. On this topic, Dan said that he power requirements and thermal envelope will not go down at all for current Prescott core CPUs, even with the socket change. The way heat is dealt with will help make platforms with the new socket cooler however.

For future CPUs, Dan told us that Intel’s desktop people are looking more at the mobile chips and how to take some tips and tricks for their own CPUs. Other websites already have reported that the first dual core CPUs will be the mobile CPU with two dies on one chip. Dan would not comment on that statement at all, so it may or may not be true. Power consumption is lower, as is the thermal envelope on mobile processors, so it’s a good start for Intel to lower these two things.

All in all, the future looks bright for computers and Intel as a whole. If there’s anything that any of our readers would like to hear about, drop an e-mail to either Nick or my self and we’ll see if we can get any answers. Once again, thanks to Dan Snyder for taking his time to talk with us.