Fastest slot 1 processor: Fastest Slot 1 /100 Mhz FSB PIII?

Fastest Slot 1 /100 Mhz FSB PIII?

JavaScript is disabled. For a better experience, please enable JavaScript in your browser before proceeding.

Status: Not open for further replies.

Jan 14, 2003

Jan 14, 2003

Add bookmark

Subject says it pretty well: What is the fastest Slot-1 PIII still on the 100 MHz FSB? Thank you oh wise gurus of the Intel CPU. — View image here: http://arstechnica.infopop.net/infopop/emoticons/icon_wink.gif — -Tom the Melaniephile

Jan 14, 2003

#2

Jan 14, 2003

Add bookmark

1ghz. Slot 1 100fsb 1ghz chip. -=Russ=-

Jan 14, 2003

#3

Jan 14, 2003

Add bookmark

Believe it or not, there was a 1Ghz Slot1 processor available and it was meant for 100Mhz FSB. They were fairly limited in supply as I understand but non-the-less, available! You can still get the latest P3 based Celeron’s which run on the 100Mhz FSB and they boast the same L2 cache (256kb ATC) as the 1Ghz Slot1. I believe the latest (and final) speed of the P3 Celeron’s is, 1.4Ghz. You will have to buy a Slocket convertor though, to use one in a Slot1 board. robtec88

Jan 14, 2003

#4

Jan 14, 2003

Add bookmark

Thanks, folks. Haven’t played with a slocket before, though I have read about ’em here. Older Dell PIII 450 machines at the radio station and I’m checking on upgrade options (throwing away the Dells is not currently an option, thanks.) -Tom the Melaniephile

Jan 14, 2003

#5

Jan 14, 2003

Add bookmark

1 GHz slot processors were just above $100 last time I checked. Assuming these are not dualies though, I’d go with Tualatin Celerons on slockets. Upgradeware adapters should be about $20, and you can pick between 1 and 1.4 GHz chips depending on budget and performance needed.

Status

Not open for further replies.

Intel’s Slot 1 CPUs Uncovered

When you purchase through links on our site, we may earn an affiliate commission. Here’s how it works.

Intel’s first Pentium II CPU with the internal product specification number ‘80522 ‘ had the codename ‘Klamath’ and is still sold as the Pentium II versions 233 up to 300 MHz. It has got four second level cache modules with the speed marking of 7 ns and the tag RAM chip 82459AB . Due to the fact that it’s only rated for speeds up to 150 MHz (300 MHz/2), the Klamath’s second level cache timing is the fastest of all Pentium II CPUs, giving it a performance edge compared to Deschutes CPUs at the same clock speed, which of course is academical, since Deschutes starts at 333 MHz in a Pentium II. The Klamath core is larger than the core of the Deschutes, because it’s still manufactured in .35 micron technology. This is also responsible for the larger heat production of the Klamath, compared to the Deschutes core. The common core voltage requirement of the Klamath is 2.8 V .

Many or most of the Klamath CPU’s have the multiplier restricted, so that you cannot exceed the allowed core frequency as long as the front side bus frequency stays unchanged. Some 266 MHz CPUs can run at 300/66 however, so that some people specialized in counterfeiting the plastic package of a Pentium II 266, selling them as Pentium II 300 CPUs . Usually Pentium II 233 and 266 CPUs come without second level cache ECC and Pentium II 300 are pretty much all sold as second level cache ECC version. Thus c’t-Magazine is asking every owner of a Pentium II 300 to check if their CPU is supplied with ECC level 2 cache, using a little program you can download from their website . If you have got a Pentium II 300 without ECC L2 cache you have most likely got a counterfeit Pentium II 266. This story is pretty valid, although I’ve once bought a Pentium II 266 in San Jose, that is not overclockable, but it has ECC L2 cache although it is only a 266 MHz version.

The Klamath is certainly overclockable, but in most cases it requires the increase of the front side bus frequency, which is dangerous for your PCI and AGP devices, because they run at PCI/AGP clocks above spec. Some crazy people overclock Pentium II 300 CPUs up to 375 MHz and even more. This is certainly an interesting thing to do, alas it is pretty useless for anyone who needs a reliable system. As a crazy overclocker you’ve got to live with the fact that your system can crash any time, if it’s screwing up your presentation for the next morning or ‘only’ a Quake 2 rocket arena deathmatch. The L2 cache, in a Pentium II running at half of the core clock frequency and rated at 7 ns is absolutely not designed for clock speeds exceeding 150 MHz, maybe it will do 166 MHz in some cases, but don’t expect any reliability of a 7 ns L2 cache running at 175 MHz or even more.

The package of the Klamath core itself shows that the chip hidden under the octagonal metal cover is larger than a Deschutes.

Please note the four second level cache modules, rated 7 ns and the tag RAM chip on the opposite side of the CPU, saying ‘82459AB’. You will not find the resistors that differentiate a 66 from a 100 MHz front side bus CPU. The Klamath is only rated for 66 MHz FSB and thus the PCB has the FSB detect Pin ‘B21’ hardwired to 0 V. Intel Pentium II 333 — 400, Core Codename Deschutes .

Intel managed shrinking the core of their 6th generation MMX CPU to 0.25 after the Klamath was on the market for roughly 8 months. This resulted in higher core frequencies, lower voltage requirements and less heat production. However, higher core speeds mean higher second level speeds as well, so that Intel needed to change the modules for the second level cache as well as its timing. The new L2 cache modules are packed at a higher density, so that only two modules are required. These new modules are rated 5.5 ns for 333 and 350 MHz Pentium II CPUs and 5 ns for 400 MHz Pentium II CPUs . The tag RAM chip has changed as well, the 333 and 350 MHz Pentium II is using the newer 82459AC tag RAM, the 400 MHz version comes with the latest and fastest 82459AD chip.

The new generation of Pentium II CPUs is now running at 100 MHz front side bus, whilst the Klamath as well as the Deschutes at 333 MHz are only supposed to run at 66 MHz FSB. Intel’s BX chipset is able to recognize which kind of CPU is plugged into the motherboard, so that most BX motherboards are following the Intel spec and supply the correct front side bus frequency automatically. The detection of the correct FSB clock is done via CPU pin B21 , which is on low in case of a 66 MHz FSB CPU and high in case of 100 MHz FSB . In a Deschutes CPU the logical state of pin B21 is determined by two SMD resistors on the CPU PCB, R5 and R6 . In case of a Pentium II 333 these two resistors are 0 Ohm , in case of a Pentium II 350 or 400 R5 is 1000 Ohm , R6 is 3,300 Ohm . (Literature: c’t-Magazine, issue 8, page 177) This may be interesting for some very courageous people who want to solder the 1 kOhm and 3.3 kOhm resistors instead of the 0 Ohm ones onto the PCB.

1

Current page:
Intel Pentium II 233 — 300 CPUs, Core Codename Klamath

Next Page From The Intel Pentium II 350/400 Data Sheet

Join the experts who read Tom’s Hardware for the inside track on enthusiast PC tech news — and have for over 25 years. We’ll send breaking news and in-depth reviews of CPUs, GPUs, AI, maker hardware and more straight to your inbox.

Contact me with news and offers from other Future brandsReceive email from us on behalf of our trusted partners or sponsors

Slot 1 vs Socket 7

Introduction

Why does Intel need new interfaces?

Is there a future for Socket 7?

What does the future hold for us?

Seven Ways to Evolve Socket 7

Intel’s New Interface Glossary

The new processor interface developed by Intel for its P6-class chips made competing manufacturers feel quite uncomfortable. The PC-compatible computer industry was divided into 2 camps, fighting among themselves for the right to determine the future architecture of the PC. As always, it was the users who came under fire first of all. Let’s try to figure out whether the RS standard, which was originally considered open, will soon turn into someone’s property (become proprietary).

The story began in 1995, when Intel introduced the first P6 class processor — Pentium Pro and a new interface — Socket 8 for its connection with the motherboard. Socket 8 has 387 pins and is not compatible with Socket 7, the standard ZIF (Zero Input Force) 296-pin socket used by all P5 class processors — Intel Pentium, AMD K5 and K6, Cyrix 6×86 and 6x86MX and Centaur Technology IDT-C6. In May of this year, Intel introduced another P6 class processor — the Pentium II and a new interface — Slot 1. From the point of view of the electrical circuit, Slot 1 is identical to Socket 8, but from the point of view of the physical implementation, Slot 1 differs significantly from previous standards.

Rather than encapsulate the processor in a small ceramic case with contact pins, Intel encased the Pentium II in a much larger plastic cartridge, which it called the Single Edge Contact (SEC) cartridge. It is a daughtercard in a protective case and requires a Slot 1 connector with 242 pins on the motherboard.

However, Intel didn’t stop there! In mid-1998, the company plans to introduce a new Pentium II processor called Deshutes and … a new interface for desktop systems — Slot 2. For Notebook computers, the company plans to release a smaller version of Slot 1, which will also work with Deshutes series processors.

However, even so many new interfaces are not a problem. The problem—according to Intel’s competitors—is that all of these new interfaces are proprietary to Intel, which has licensed their use to motherboard manufacturers but doesn’t want to license their use to competing x86-compatible processor manufacturers.

As a result, AMD, Cyrix and Centaur cannot manufacture processors that would work on Slot 1 motherboards, chipset manufacturers can only support Slot 1 if they have been licensed by Intel. Motherboard manufacturers cannot release boards that support any P6 class processor, as was the case with Socket 7 and P5 class processors. They also can’t produce Socket 7 boards on new chipsets from Intel, since they no longer support it, and are significantly limited in the choice of chipsets for boards with Slot 1, since some possible suppliers did not get access to Intel technology.

In short, everyone was in trouble. Motherboard manufacturers must make a choice: who to be with? Users, in turn, also have to choose what to buy. At first glance, the choice is obvious, as Intel processors outperform the competition. But don’t forget that Intel products cost an average of 25% more than competitors’ products, and motherboards that support SEC cost $30-$100 more than motherboards with Socket 7.

The struggle in the PC market has always been in its favor. However, now the situation looks as if the PC is going to follow the example of the Macintosh — an architecture that was originally «closed».

Why does Intel need new interfaces?

The easiest way to scold Intel. «Intel is trying to do exactly what IBM did in 1987 by trying to implement its own Micro Channel bus,» says Glen Henry, president of Centaur Technology, formerly of IBM. «Actually, that’s one of the reasons I left IBM. I thought they were being stupid.»

Henry is referring to the Micro Channel architecture, an external device bus that IBM offered as a replacement for the ISA bus. The architecture was owned by IBM, and other companies had to obtain licenses from it. If successful, the Micro Channel architecture could force a number of companies out of the market and give back to IBM control of the PC architecture that had been lost in 1981. However, the idea failed.

Without a doubt, Intel has partially taken control of the PC architecture. Indeed, the most important components of a PC are the processor, system chipset, graphics controller, memory chips, and motherboard. Intel produces about 90% processors, 80-85% system chipsets, and is the largest motherboard manufacturer. Intel recently announced its intention to acquire Chips & Technologies (antitrust clearance still pending), which makes components for computers like the Notebook and has been working with Intel and Hughes to develop a 3D graphics controller chip (codenamed Auburn). It is planned that the controller will compete with products from companies such as S3. The company also owns a stake in Rambus, a company that has a good chance of setting new standards for memory chips. Intel and Rambus are introducing Rambus DRAM (RDRAM) as the next step after Synchronous DRAM (SDRAM). If successful, this will be the first time since 1974 years of one company controlling DRAM standards.

Surprisingly, Intel has achieved such success without selling a single computer under its brand name. It seems that the only area where Intel has not reached is the operating systems, which remain the domain of Microsoft. However, is Intel really an «evil empire»? When answering this question, keep in mind the following:

Like any other company, Intel has the right to develop technologies and protect them with patents. «As an Intel shareholder, I would be very upset if the company gave away its most significant secrets for free,» says Manny Vara, public relations manager for Intel’s desktop processors.

Intel had good reasons for creating a new processor interface. Socket 7 does not provide sufficient bus bandwidth, especially in multiprocessor (MP) configurations. All communication in the Socket 7 architecture goes over a single 64-bit bus. With a standard bus frequency of 66.6 MHz, the maximum throughput is about 533 MBps (megabits per second). Even if next year the bus frequency is increased to 100 MHz, the bandwidth will still be insufficient for high-end systems.
So Intel added a second 64-bit bus to the Pentium Pro and Pentium II processors. This additional bus (backside bus) is designed to exchange with the second level cache (L2 cache) and is faster than the external bus. Such an architecture requires at least 72 more contacts in addition to those available in Socket 7.

Not everything is so gloomy. Intel gives others access to some of its new technologies. For example, she has helped companies such as Corollary, NCR and Hyundai develop P6 compatible chipsets for multiprocessor systems. Intel stated that these companies are «promoting the evolution of the PC architecture.»

Intel cannot take ownership of the PC architecture because it was not originally an open architecture. This goes against the conventional wisdom that PCs are more popular than Macs because the PC architecture is open and cloneable by anyone. In fact, both the IBM PC and the Mac were originally closed standards. PCs were cloned by everyone who felt like it, simply because it was easier to clone them. By developing a bus protocol that is difficult to clone, Intel proved once again that the PC is not an open architecture.

Does Socket 7 have a future?

It may seem that all of the above means the decline of Socket 7 and Intel’s competitors. But it’s not. While Socket 7 doesn’t provide the bandwidth needed for high-end systems, it’s good enough for entry-level and mid-range computers, which are the kind of computers that most users buy. Capturing this market is becoming a major challenge for AMD, Cyrix and Centaur.

«We are not limited by Socket 7. I want to tell everyone this,» says Lance Smith, Technical Marketing Director of AMD, «Most of all we are limited by time — the time we need for the release of a new product.

Early next year, new processors, system chipsets and SDRAM chips will bring the Socket 7 bus frequency up to 100 MHz. This will increase the throughput by 50% — up to 800 MBps. At the same time, developers are trying to reduce the exchange on the bus. Increasing the size of the L1 cache increases the likelihood that the processor will find the required instructions and data without going to the bus. At the moment, all new processors using Socket 7 have L1 cache increased to 64 KB. This is 2 times more than in Intel processors — Pentium, Pentium Pro and Pentium II.

It looks like AMD is going to increase the size of the L1 cache even more after moving to 0.25 micron technology in the near future. The new technology will reduce the die area from 162 mm ² to 68 mm ² and thereby free up space for additional cache while maintaining cost.

Centaur follows the same path. At the end of next year, the company plans to introduce an improved version of the IDT-C6 chip with an integrated L2 cache — a solution found in some RISC chips, but first used in the x86 family of chips.

What does the future hold for us?

Increasing the size of the built-in cache is not the only means in the struggle for the survival of Socket 7. There are several other ways to develop.

One way is to design processors with an additional L2 cache bus, just like Intel does. Instead of developing a new socket with an increased number of pins, the processor and L2 cache can be placed on a daughter card inserted into Socket 7. The additional bus can then operate both at the processor frequency and at some fractional frequency — most likely at half the processor frequency — for reducing the cost of SRAM. The external bus runs at 66.6 or 100 MHz to ensure backwards compatibility.

The same can be done even better by placing the processor die and L2 cache in the same package, called a Multichip Module (MCM), just like in the Pentium II processor. The processor and L2 cache communicate over a fast auxiliary bus, and the MCM cartridge is installed in Socket 7. MCM production is quite expensive, but AMD uses IBM’s C4 technology, which reduces costs. This technology allows you to place the leads to which the contacts are soldered on the entire surface of the crystal, and not just at the edges. NexGen, acquired by AMD, did something similar 2 years ago: one version of the Nx586 processor contained a separate floating point chip in the MCM. This path is also possible (but less attractive) for Cyrix, whose processors are manufactured at IBM factories.

Another way is to place the in-line or look-through L2 cache on a child card installed in Socket 7. In this case, there is no need for an additional processor bus. Instead, an external clock generator sets the processor to a higher frequency. The exchange of the processor with the L2 cache takes place at this increased frequency, but the cache controller exchanges with Socket 7 still at 66.6 or 100 MHz. This way allows you to abandon the development of new processors.

«Either of these alternatives will extend the life of Socket 7 and achieve the performance of Slot 1,» says Mark Bluhm, vice president of strategic planning at Cyrix, «this socket has at least a year or two of life ahead of it «.

Seven Paths of Socket 7 Development
900 99 1. Increasing the frequency of the processor bus

Path Strengths Disadvantages Intended Developers

Increases the performance of the L2 cache and main memory Increases the cost of motherboards. Presumably narrows choice of motherboard vendors All competitors Intel

2. Increasing L1 cache size Increasing cache hit probability Increasing processor die area AMD

3. Increasing the size of the L2 cache Improve cache hit chance Requires more SRAM chips Unknown

4. Add extra bus for L2 cache in Socket 7 Unknown

5. L2 cache integration Significantly speeds up access to the L2 cache, eliminates the need for an external L2 cache, provides the ability to install an L3 cache increases the processor crystal Centaur, AMD

6. Using the in-line cache L2 accelerates access to the L2 cache, without increasing the processor crystal complicates the processor architecture; requires a daughter card installed in Socket 7 System chipset manufacturers

7. Multichip module (MCM) L2 cache placement Speeds up L2 cache access using an additional bus. Does not require a daughter card Increases processor cost if IBM C4 technology is not used AMD, Cyrix

However, long term solutions are needed. Two options are possible: development of motherboards with a universal slot, in which daughter cards with any processor would be installed, or the creation of a new connector that directly competes with Intel’s developments.

The first option is already being implemented. At the Computex show in Taipei in June, Asus received the BYTE Best of Computex award for the P/I-P65UP8 motherboard with a special socket for processor daughter cards. The daughter card can have a Socket 7, Socket 8 or Slot 1 processor along with the corresponding chipset. At the moment, this connector is patented, but if enough companies agree on a standard, any company can produce universal motherboards. The disadvantage of this design is the presence of an additional connector between the processor and the motherboard, which introduces noise and signal delays at high clock frequencies.

Another way is to create a new processor interface without using Intel patents. None of the competing companies are open to talks about this, but even if the interface is developed, there will still be a BIG problem: Intel. The new interface will not be supported by the largest processor manufacturer (Intel), the largest motherboard manufacturer (Intel), and the largest system chipset manufacturer (Intel). The result could be two PC architectures, one controlled by Intel and the other supported by a majority of vendors controlling only a small fraction of the market.

Glossary of new interfaces Intel

Socket 8 — ZIF (Zero Input Force) 387-pin socket for the Pentium Pro processor. It requires a multichip module, a special package that includes a processor die and one or two SRAM dies for the L2 cache (from 256 KB to 1 MB). The L2 cache can operate at the processor frequency (currently up to 200 MHz). The Pentium Pro supports cached memory up to 64 GB and is used in high performance workstations and servers.

Slot 1 is a 242-pin daughter card slot for a Pentium II processor housed in a Single Edge Contact (SEC) cartridge. Inside the cartridge there is a processor die and SRAM chips (up to 512 KB). Currently cacheable memory is up to 512 MB, which is much less than the Pentium Pro. One or two Slot 1 connectors can be installed on the motherboard. The standard external bus frequency at the moment is 66.6 MHz. It is planned to increase it to 100 MHz with the release of the 440 BX chipset in the first half of 1998 years. The L2 cache can operate at 1, 1/2, or 1/3 of the processor frequency. For now, 1/2 is the standard, as higher frequencies require expensive SRAM chips, and Intel targets the Pentium II for midrange workstations and servers.

Slot 2 is a new daughter card slot designed to accommodate a Pentium II processor in a larger cartridge. Slot 2 is not intended to replace Slot 1, but is planned for top-tier workstations and servers. Up to 4 such connectors can be installed on the motherboard, and more when using special chipsets. The extra volume of the cartridge is used to install SRAM chips. The processors installed in Slot 2 support cached memory up to 64 GB. With the use of expensive burst SRAM, the L2 cache can run at the processor frequency. The external bus frequency is at least 100 MHz.

Intel’s mobile slot — doesn’t have a nice name yet. Represents a smaller copy of Slot 1 for computers such as Notebook. Completely identical to Slot 1.

Contrary to some reports, future P6 class processors from Intel will not require special sockets. So, Deshutes is a Pentium II, made by 0.25 micron technology. Versions for Slot 1, Slot 2 and Mobile slot will be released.

Based on information from BYTE magazine and Centaur Technology, AMD, Cyrix, and Intel

Pentium Pro for the home. Remote professional / Habr

Among the limited set of holy grails of the modern retrofan, there is one from which, contrary to the requirements of sanitation, probably all more or less popular retro reviewers have already drunk. And some more than once.

Is there life on Venus?

So, Pentium Pro. The most professional, the best, the most large mid 9 processor0x. Year 1995. The sixth generation of Intel processors, an unprecedented calculator of the «money no object» class (English «money does not matter») . The price of completed assemblies in the region of five thousand bucks is not shocking, except perhaps for a corporate client who was just looking for a new server in the office for 200 tables.

In 1996, Intel made an attempt to push the Pentium Pro into the relatively affordable workstation segment. The i440FX chipset was released and the motherboard on it is the VS440FX «Venus». Venus. It looks erotic… ATX board with PS/2, USB 1.0 and DMA IDE modes, as many as four PCI 2.1 slots, which means there is a Bus Master. In general, at first glance, quite an ordinary motherboard … for 19 years98!

Intel 440FX is a two-chip solution, not counting the southbridge.

Of course, for an advanced retrofan, it will not be difficult to see “senile wrinkles” on this Venus in the form of SIMM memory and . .. the absence of AGP. Not being a direct collector of PCI video cards, I chose from only three options:

Install a two-megabyte S3 Trio64 and play a serious uncle who has no time for games at all.

Install a 4MB S3 ViRGE and run the original Tomb Raider 1996 years in S3D rendering.

Install Nvidia Quadro NVS PCI, similar to GeForce 4 MX, and have no limitations in graphics tests at all.

Nvidia Quadro NVS PCI. Chip NV17.

But the main character today is still a processor. It is superscalar, three-pipeline, out-of-order, and a completely new instruction decoder for confident code parallelization. Full-speed cache up to 1MB, 36-bit PAE with theoretical support for up to 64GB of RAM. In general, the brain of even a healthy, prepared pekboyar could easily detonate. After all, at 19In 1996, assemblies based on Socket-3 (486) were still being sold with might and main, AMD K5 occupied the middle segment, and the first Pentium was de facto a top product, because Pentium Pro was far from being found in every metropolitan store.

PentiumPro 200MHz, 256kB cache, Socket 8 (387).

The copy I got has the latest stepping, which fixed quite serious errors in the kernel. But at the first start of the system, I still flashed the BIOS to the latest version. Habit! And then I installed Windows 98 SE… Yes, I also read the Wikipedia article, and I remember that the Pentium Pro has problems with 16-bit code execution. But here we must not forget that the current OS in 1995 was Windows 95, sorry. And it really had a lot of 16-bit modules, which led to a decrease in performance on the Pentium Pro.

How ironic! After all, 16-bit libraries were designed to speed up the work of a heavy operating system for a typical computer of that time! It’s just that the P6 architecture, suddenly, turned out to be is too 32-bit … But, anyway, in 98SE the situation is already a little different, because the Win9x development team gradually tried to abandon 16-bit pieces, although it often ran into the volume of home PC RAM.

There is a fortune on this screen. And not only by the standards of 1996!

The amount of RAM we have is not very huge either: 64MB. And if you don’t bother searching for Avito, then today for four SIMMs this is practically the limit. However, this did not prevent us from conducting a fairly detailed testing of this amazing system. And the main competitors today will be Pentium II 266, Celeron 266 and Celeron 300A. All of them will be tested on an AOpen AX6LC motherboard based on the i440LX chipset, with the same amount of memory and the same video card. Various assemblies from my old reviews will also appear in the charts.

The opponents are very serious, even if none of them has the title of «Pro»…

Performance at a professional level

CPU-Z Vintage Edition shows us how much faster the Pentium Pro is than the Pentium MMX at the same frequency: a 1.5x difference! It’s a pity that our «professional» has nothing against Pentium II and even Celeron without cache . Perhaps this is how it should be? After all, these processors are direct successors to the P6 line. Yes, and the same VIA Cyrix III, which I dared to put it next to the Pentium Pro — it lost to the old man in the FPU test, and quite decently. The increase from overclocking to 233 MHz is not shocking, but it is there, scalability is present.

In WinBench 99, the closest results to the Pentium Pro were obtained by its descendants — Celeron 266 and 300A. Moreover, both are ahead in the FPU subtest, but in integer arithmetic, 266 lost a lot! It turns out the opposite result compared to CPU-Z. But the average level for this «retro hospital» is already clear. The comparison with VIA C7 is sobering, even if there is a whole decade between them. In Windows 2000, the CPUMark subtest gives some very strange results, and scalability in overclocking has gone somewhere …

On the other hand, in SiSoft Sandra arithmetic, the advanced Pentium Pro architecture makes it possible to keep up with VIA C3 at a four times higher frequency: 200 MHz versus 800 MHz. It’s not a big victory, of course, but what a victory it is — against the newer Intel processors, even the overclocked Pentium Pro 233 frankly does not pull. The difference between Windows 98SE and 2000 this time is within the margin of error.

But in the Multimedia test, everything is more interesting. The Pentium Pro at 233MHz almost lost the PII-266, though only in one subtest — the FPU. And it seemed to me that this is the same bench, where if there is no MMX, there is no chance! However, unexpectedly.

But it’s all boring compared to SuperPI’s fun pi counting. Windows 2000, where are you ?! They just changed the operating system, and instead of an uncertain victory over the Pentium MMX, it is already possible to compete with the Pentium II, and even with the Celeron on the advanced Mendocino architecture, not to mention its frequency.

Would you believe it the first time?

When counting 1 million characters, in principle, everything is the same. The fast PPro cache eliminates the difference in frequency and memory speed on the i440FX platform. Yes, and the gaps here reach minutes , which is simply unbelievable. For example, I didn’t believe it. But repeated tests were relentless…

In WinRar, the numbers are much more mundane, but at least Celeron 266 is far behind. And in general, we should be glad that the Pentium II surpassed its pretentious ancestor so much, and then the people’s Celeron easily jumped over it.

Beautiful, clear, and absolutely comparable!

Mandatory program

3DMark9 graphics test9, and — all the main competitors in more or less the same conditions. The miracle here seems to have happened , but it is very inconspicuous. Pentium Pro, despite the difference in frequency and age, outperformed Celeron without cache. There are about three years between copies and there is a whole gap in market positioning.

A motley company on backing vocals here shows what 3DMark99 itself is capable of — is comparable to systems? Only the assembly on the VIA processor can be taken seriously. Once again. And while it’s clear that developing a processor, even a garage-built one like the Cyrix III, is a complex and multifaceted process, wasn’t the Pentium Pro the starting point for Centaur in terms of consumer performance? (spoiler: no; they tried to get Mendocino through IPC).

3DMark 99 MAX: The Pentium Pro still holds up well. Visually, the tests are not too slow

3DMark2000 does not start on Pentium Pro because it needs the MMX instruction set. But 3DMark2001 doesn’t need them! Saying goodbye to the remnants of logic, we analyze the numbers. Of course, our hero cannot reveal , of the fourth GeForce on PCI, but over the years of collecting various hardware, I have recorded a lot of comparable figures. For example, I was impressed by the attempt of Celeron (Coppermine) 667 to get a formidable Pentium Pro with a formidable Nvidia Quadro on a sad Intel-built-in. The Pentium II and Celeron 300 have traditionally disappeared beyond the event horizon, and the Pentium Pro at 233 MHz has never passed this test — the availability of overclocking (usual jumpers here) does not guarantee its stability.

DroneZ also turned out to be too tough for the “accelerated” assembly, and Windows 2000 failed. Celeron on the Mendcino core is once again out of competition. Covington also tried, but at the same frequency, he would have lost. It is unlikely, of course, that any of the owners of the Celeron-266 was worried, because for the cost of one Pentium Pro you can buy a whole box of 266, a couple of which will start at a frequency of 400 MHz. But I’m already trying to draw conclusions. But I have another test up my sleeve…

So Quake. Everything is here at once: both the soft mode, and OpenGL, and Windows 98, and 2000 — the results of which would be better to unsee altogether. Although it would seem that OpenGL and Windows 2000 are friends forever! In general, this is the essence of one phenomenon, like a mouse and a pad, like a floppy disk and a blue screen. But, as you can see, not everything is so simple, and it was not for nothing that Windows 98 remained a gaming operating system both during the time of Windows 2000 and after the release of XP. Quake itself is, of course, a product of its era, even though I used WinQuake for soft runs. It can be seen that our «professional» is hard here. You can see it in tests 1996-1997, by the way, so we did not make any discovery here. Left Covington behind — already worthy.

Pentium Pro 200 surprises in OpenGL! But only in Windows 98…

Well, at the very bottom of the graph, two processors from the mature Socket 7 times — Pentium 200 MMX and Rise mP6 with a similar frequency, modestly joined. And now it is clear that it was them that PPro was supposed to surpass, and brilliantly coped with its task. Everything that happened after is a consequence of his achievements, and nothing else.

Points for speed and artistry

Before turning to conclusions about the Pentium Pro, let’s pay attention to the elephant in the room: the Pentium II does not hold up well against the Celeron on the Mendocino core. Previously, I did not arrange such tests: even knowing the result, what kind of retrofan would put a budget processor in his assembly instead of an elite Pentium II? But the graphs won’t lie either — on this historical canvas, the Pentium II looks like is already outdated .

On the other hand, I haven’t run any low-level cache tests. Indeed, in order to obtain reliable, comparable results, you need to at least install a Pentium Pro on a 440LX (and there are such methods), but right now, groping for some conclusions in the dark, I come to a paradoxical one — cache speed was not as important as its integration to the core!

What different approaches! What a flight of engineering thought! And users look only at «MHz»…

From this angle, the Pentium Pro can be considered a model of the future Celeron. And although the goal for the budget design was just a compact full-speed cache, the bonuses from the elimination of additional chips and buses actually played a bigger role, which certainly came as a surprise to both Intel itself and the market. Perhaps, Intel should have presented Mendocino as a Pentium III, along the way removing the stupid 66MHz limit on the system bus. True, then ordinary users at 90x would never get fast and inexpensive builds.