Intel Roadmap Update June 1999 — Part I, The Highlights — THG.RU

Introduction

You haven’t found any roadmap updates on this website for quite a while. The reason is pretty simple, I didn’t really consider any of the changes in the last six months really interesting enough, the main track never really changed anyhow. The latest update consists of a quite impressive number of newsworthy stuff though, so that I finally decided to make an article out of it.

There are obviously a lot of new things coming up, as you will see in the full overview published later on. Before we get into this dry material I’d prefer to present some of the highlights of the latest roadmap.

The first thing that caught my eye was this interesting quote on a page called

550MHz Pentium III Processor Press Feedback

This incredible quote that makes you feel as if your kids might need some parental guidance before reading it doesn’t just come from somewhere, it’s from a supposedly ‘independent’ hardware website. It seems clear that Intel did a great marketing job with this particular editor and I wonder if this person thinks he’s indispensable too, or if he was just too intoxicated with some different stuff before he wrote those lines.

Pentium III Goes Socket370

There was quite a bit of speculation in the past about this logical move and now it’s finally confirmed, Pentium III (new .18µ ‘Coppermine’-core) will go ‘FC-PGA370’ (= Flip Chip Pin Grid Array 370) by Q4/99.

Pentium III for Socket370 will only be certified for uni-processor systems and initially there won’t be any Coppermine-variants with 133 MHz front side bus available for Socket370 and the supply will be limited vs. the supply of Coppermine for Slot1. This will change in the second quarter of 2000, when supposedly all Coppermine-Pentium III variants will be available for Socket370 as well as Slot1.

Possible New Socket418 for Dual-Pentium III Configurations

Intel is currently ‘exploring’ a new Socket called ‘FC-PGA418’ for dual-processor systems. This Socket is supposed to be backwards compatible to Socket370. If Intel should decide to go for Socket418, we can expect it in the second quarter 2000, when Pentium III will be available in all speed variants for Slot1, Socket370 and the new Socket418. It’s not quite understandable why dual-processing would require 48 pins more, since as we know Celeron runs fine in dual-mode and Socket370.

Pentium III for Notebooks Will Ship Soon

It won’t take long anymore and Pentium II for notebooks will be old news too. The mobile Pentium III will start shipping at clock speeds from 450 to 600 MHz in the third quarter of this year. These mobile PIIIs will basically be the same as Coppermine, manufactured in .18µ-tecnology, with 256 kB on-die L2-cache, but the version running at 600 MHz will include the so called ‘Geyserville’-technology, which clocks down the CPU as soon as it’s running from the battery, the core-voltage will be lowered and thus the power-consumption will come down to make the battery last longer. All the new mobile-PIIIs will actually run at 100 MHz bus speed.

Intel doesn’t give up on the 3D-graphics market

As if the performance of i752 wasn’t sad enough, Intel will soon present i754, the AGP4x-version of i752, which will be launched together with the upcoming i820-chipset (‘Camino’). The plan continues with ‘Capitola’ and later on ‘Capitola MLK’ for the ‘professional/enthusiast’-segment, starting early 2000 and ‘Indian Beach’ for the ‘performance’-segment, expected at the begin of Q2 2000. We shouldn’t expect too much of it, since Intel is currently counting i752 into the ‘performance’-segment too, although its performance ranges below NVIDIA’s good old RIVA TNT.

The powerful Pentium III Xeon for servers will stay at 100 MHz FSB

Here we’ve got another example of Intel’s beautiful logic. Whilst PIII will move to 133 MHz front side bus with the introduction of the 820-chipset (‘Camino’) in September, the large L2-cache variants of PIII Xeon will stick to 100 MHz FSB until the introduction of Merced in the second half of 2000. The PIII Xeons with on-die 256kB L2-cache (codename ‘Cascades’) will be launched together with the workstation and server chipset i840 by the end of Q3 1999, and they will run at 133 MHz front side bus. This is another proof that Intel is trying to fool us a little bit. We are supposed to learn that RDRAM and its higher memory bandwidth is really essential for the future, but the CPUs used in servers, where the memory load is the highest, can only transfer data over the good old 100 MHz FSB. So Intel wants to tell us that business software is in more urgent need for a higher data transfer rate than server environments? Come on guys, give me a break, will ya?

Intel has no plans of implementing PC133 SDRAM

Intel does not believe into PC133. I was told that Intel feels that PC133 is not offering enough performance gain and that it’s unreliable on top of it. The protocol overhead of SDRAM is indeed a significant problem, which does not get reduced enough even though you’re finally transferring the actual data 33% faster. One of the main problems of PC133 is that PC133 with a CAS-latency of 3 is performing worse than PC100with a CAS-latency of 2, but there are several other issues as well, which will become even more significant with DDR-SDRAM and DDR2-SDRAM as we move ahead.

The alternative explanation could be however, that Intel really despises PC133 because the competitors in the chipset market are using it as their weapon against RDRAM. Whichever it may be, PC133 will not be used by Intel as found in the roadmap, even upcoming chipsets that don’t work with RDRAM will only use PC100 SDRAM.

The Desktop CPU Roadmap will be published later on today or tomorrow, followed by the Mobile CPU Roadmap, the Workstation/Server Roadmap and the Pricing Roadmap.