Community
Discuss & share your modifying mayhem with like minded enthusiasts.

Intel Pentium III 1GHz - Anandtech 08/03/2000

  • Martin
  • Martin's Avatar Topic Author
  • Offline
  • Administrator
  • Administrator
More
7 months 21 hours ago - 7 months 21 hours ago #1 by Martin
on the 08/March/2000 
File Attachment:
 published this review of the Intel Pentium III 1GHz by Anand Lal Shimpi

 

Intro
The switch over to Intel’s 0.18-micron Coppermine core has occurred very quickly.  It’s hard to believe that five months ago the Coppermine didn’t even exist and AMD’s Athlon was on its way to completely dominating the performance desktop processor segment.  In comparison to the older Pentium III based on the 0.25-micron Katmai core, the Athlon had no problem completely wiping the floor with the chip. On a clock for clock basis, as well as on an uneven playing field, the Athlon ended up ahead of the Pentium III in every single test we could throw at it including some tests specifically optimized for the Pentium III’s SSE instructions.  Things looked very promising for AMD and their newly born Athlon, and at the same time, things weren’t looking good at all for Intel who had had a very easy time competing with the K6-X line of processors from AMD.Ideally, Intel was to release their new Pentium III with a full speed on-die L2 cache as well as a number of other enhancements on October 25, 1999 alongside their new desktop chipset platform, the i820.  Unfortunately, the launch didn’t go exactly as planned, and although the new Pentium III ended up making its way out the door on that day, it’s partner, the i820, didn’t make it.  It wouldn’t be until later in Q4 when the i820 chipset would actually see the light of day, and even then it was plagued with a relatively poor welcoming from motherboard manufacturers simply because the incredibly high costs of RDRAM, the i820’s only natively supported memory type, kept it from being a viable alternative to the BX platform. 

 In addition to the lack of a good platform to run the processor on at its launch, Intel was plagued by supply issues which resulted in a complete lack of Pentium III CPUs in the usual channels for the entire month of February of this year.  According to Intel, these problems weren’t related to yield, but rather a simple case of the demand exceeding the supply of parts.  The processors that were available didn’t exhibit any signs of a poor yield, especially considering the success we, as well as many other users, enjoyed with overclocking the 500E and 550E processors to speeds of 750 and 825MHz in some cases. The combination of an unattractive motherboard platform with poor availability left Intel in a position they’re not used to, and gave AMD quite a big opportunity.  AMD exploited that opportunity earlier this week with the release of the world’s first air-cooled x86 processor operating at a frequency of 1000MHz, or 1GHz.  Luckily, Intel was able to respond to AMD’s recent processor launch with a release of their own. It’s time to welcome the newest addition to the Pentium III family, just 5 months after the release of the 733MHz parts, the 1GHz CPUs are finally here.

Pentium III 1GHz Specifications

.         29 million transistor 0.18-micron Coppermine core
·        1GHz clock speed – 7.5x clock multiplier
·        32KB on-die L1 cache running at core speed
 ·        256-bit Advanced Transfer Cache - 256KB on-die L2 cache running at core speed
·        Advanced System Buffering
·        242-pin Slot-1 GTL+ CPU interface running at 133MHz
·        1.70v core voltage

For starters, the 1GHz Pentium III isn’t any different from the original Coppermine based Pentium III that was released last October -- with one notable exception.  Like AMD, Intel has resorted to increasing the core voltage to 1.70v from 1.65v in order to retain high enough yields on the CPUs at this frequency.  The 1.70v core voltage is still within the range of tolerance for the CPU, so it is acceptable for Intel to do this. This is a practice that most overclockers use in order to increase the stability of an overclocked system.  However, when Intel or AMD does it, it’s “increasing the yield”   Another interesting thing to note here is that the 1GHz product will be available only in a 133MHz FSB version, so no 100MHz x 10.0 configurations on your old BX setup; you’re slowly being forced to get rid of that BX board.

If you’re not familiar with some of the features the Coppermine core offers, the two biggest and most talked about benefits of this core are the Advanced Transfer Cache (ATC) and the Advanced System Buffering (ATB). The ATC on the Pentium III is just the fancy name for the on-die 256KB L2 cache.  Now keep in mind that the Pentium III isn’t just a Celeron with twice as much cache and SSE, the L2 cache bus has been increased from the 64-bit bus width on the older Pentium III and Celeron processors to 256-bits wide.  The ATC also refers to the 8-way associativity of the 256KB L2 cache on the Coppermine (compared to the 4-way associative L2 cache on the old Pentium III/Celeron). Because the 256KB of L2 cache is on-die, the transistor count of the Pentium III is increased tremendously over the 9.5 million transistors that made up the original Pentium III (Katmai) core.  The 256KB L2 adds about 19 to 20 million transistors, putting the total transistor count of the Coppermine at approximately 29 million transistors. Advanced System Buffering is a simple term that represents the increase in buffers the Pentium III Coppermine offers over the previous generation of processors, including the Katmai based Pentium IIIs.  If you are interested in specifics, there are now 6 fill buffers, 8 bus queue entries, and 4 writeback buffers (up from 4, 4, and 1 respectively).  These three optimizations all help to take advantage of the 1.06GB/s bandwidth offered by the 133MHz FSB. 

The many flavors of Coppermine
Before we dive into the 1GHz chip itself, we’re going to take a quick look at the processors available from Intel today and their respective names in order to clear up any confusion:

 

 Intel’s Hurdles - RDRAM
Intel has had to overcome quite a bit in the past 6 months alone.  Not only have they had to face fierce competition from AMD, but they have also had problems with promoting the acceptance of RDRAM as a memory standard, thus creating problems for their 820 chipset which depends on the success of RDRAM in order to become a widely adopted platform.While it may seem as if RDRAM is the root of all evil and the absolute cause of Intel’s problems in recent history, you have to understand that RDRAM does have quite a bit of potential for success.  The two factors holding back RDRAM right now are the costs of manufacturing and the low yields that manufacturers that are actually producing chips/modules have been seeing. 

While there are supposedly new techniques in development and on the verge of implementation that are designed to cut the manufacturing costs of RDRAM, there is still quite a bit of skepticism from the industry about exactly how viable of an option RDRAM is right now.Yields have been increasing, and while the yields on PC800 RDRAM isn’t even remotely close to what manufacturers can claim for their PC100/PC133 SDRAM chips, it is increasing to an almost reasonable point.  One of the biggest problems is that Samsung is currently the only major producer of RDRAM in the world, and whenever there is a lack of competition in a particular market there is definitely a lack of competitive pricing. Intel specifically addressed the issues related to RDRAM at their latest Developer Forum in a track dedicated to memory technologies of the future.  In this presentation track, Intel described some of the corrective actions they are presently taking to make RDRAM a more viable memory option in 2000. Intel is currently working on adding more details to the layout guidelines in an indirect effort to help increase the yields and reliability of these modules at higher frequencies, especially PC800.  They are also implementing an enhanced validation process in an effort to bring RDRAM to the point that it is as easy to manufacture as SDRAM is right now.  Although this is quite some time away, it’s steadily approaching. According to Intel, by the end of this quarter there should be about four vendors shipping RDRAM; unfortunately, they failed to mention in what quantities these vendors would be shipping their RIMM modules.  If yields continue to be as low as they are now, even the six vendors that are due to be shipping by the end of Q2 will have trouble producing enough RDRAM in order to drive costs down.  Let’s hope, for Intel’s sake, that things do change because they recently reaffirmed their dedication to RDRAM and plan on keeping it in all of their future platform designs.The reasoning behind their dedication to RDRAM is actually quite valid.  Not only does RDRAM offer greater memory bandwidth than competing SDRAM solutions out today, it is supposedly more efficient than DDR SDRAM and features a lower pin count than DDR SDRAM as well.  A recent trend we’ve been noticing is towards serialization, and lowering pin count which allows for an easier time when doubling bandwidth as well as producing high frequency devices.  RDRAM definitely follows this trend, as do technologies such as Serial ATA, so there is definitely an opportunity for success but as of now, that potential is limited by other factors such as manufacturing costs.

Intel’s Hurdles – Reaching 1GHz
On the day of the launch of the 1GHz Athlon, AMD’s lead partners, Compaq and Gateway, began offering systems based on the newly released CPU.  The Gateway Select 1000 and the Compaq Presario 5900Z will be available with a 1GHz Athlon CPU for under $3000 effective, if not immediately, then by the time this publication reaches your screen.  How does Intel plan on responding? Three of Intel’s premier partners, Dell, HP, and IBM, will be releasing systems based on the Pentium III clocked at 1GHz with the announcement of the CPU, and like systems based on the 1GHz Athlon, these systems will be available in limited quantities, but they will be available for purchase. Now, here’s where things get weird: because Intel pushed for the early announcement of their 1GHz part and didn’t want to disturb the rest of their roadmap, they currently have no CPUs faster than 800MHz but slower than 1GHz.  The 8xx and 9xx chips are due out in a couple of weeks, but until then the next option above a Pentium III 800 from Intel is a Pentium III 1GHz. 1

 In terms of general availability, Intel is sticking to their roadmap for general availability of the 1GHz parts.  You won’t see any 1GHz parts, outside of those included in systems by Dell, HP, and IBM, until Q3 of this year, meaning that AMD will most likely get a head start on Intel in terms of shipping 1GHz parts to the general public (that includes all of us). As we mentioned at the start of this review, Intel had to increase the core voltage of the Pentium III by about 3% in order to maintain high enough yields on the 1GHz parts.  AMD had to do the same thing with their 1GHz Athlon which now runs at 1.80v instead of the 1.70v that the Athlon 850 ran at or the 1.60v that all other Athlons operate at. In addition to increasing the core voltage from 1.65v to 1.70v, Intel also had to take care of the additional heat generated by the Pentium III running at 1GHz.  Dell, HP, and IBM will be using noticeably larger heatsinks in their 1GHz boxes, an example of which is the heatsink that our 1GHz sample was outfitted with.

 
 
The heatsink featured a copper base and an increased fin surface area from what we’re used to seeing on heatsinks Intel uses on their CPUs.  AMD did the same thing with the 1GHz Athlon sample they provided us.  The Athlon at 1GHz is already producing 50W of heat, and while the 1GHz Pentium III should be lower than that, we are still talking about a considerable amount of heat.

  

 


Motherboard Support
We mentioned, and you probably already know, that the i820 motherboard platform isn’t exactly the most popular platform in the market right now.  So what kind of support does the Pentium III have? Fortunately, VIA’s Apollo Pro 133A chipset is finally performing like it should and there are more than enough motherboards based on the chipset to provide you with a few good options when picking up a board. In particular, we used the ASUS P3V4X in our 133A tests with the 1GHz Pentium III simply because our Gigabyte 133A board failed to properly support the CPU without an updated BIOS that was, at the time of testing, unavailable.  If you do recall from our  VIA Apollo Pro 133/133A Motherboard Roundup for February 2000 , the ASUS P3V4X performed at the top of its class under all of the business and content creation tests, but it dropped down to last place in the gaming tests.

  

Apparently, there are issues with the P3V4X and the VIA AGP GART driver v4.00 that cause the gaming performance of the setup to drop considerably.  The only solution for now is to use v3.59 of the GART drivers from VIA, which completely corrects the performance issues.  We are still awaiting an explanation for this anomaly, which is present only with the P3V4X, from ASUS, and when we do receive an official statement, we will let you know. In order to illustrate the maximum performance of this setup, we used v3.59 of the GART drivers with the P3V4X test bed.

Overclocking
While the 1GHz Athlon couldn’t be overclocked too far because it is simply out of valid clock multipliers, the 1GHz Pentium III wouldn’t make it too far above 1000MHz because of cooling/yield issues.  Remember that Intel isn’t mass producing these 1GHz chips, so they can’t be expected to run at the usual 20 – 50% speed increase that we normally run the Pentium IIIs. 

The Test
 

 


Content Creation performance
 
In terms of Content Creation performance, the Pentium III at 1GHz on the i820 platform is just as fast as the 1GHz Athlon on a KX133 setup. This makes sense because of the close proximity the Athlon has held to the Pentium III, even at the higher clock speeds where the Pentium III's full speed L2 cache should theoretically cause the CPU to begin to pull away from the Athlon. A possible explanation for this is that the Content Creation tests are benefited by the Athlon's 64KB data cache that runs at its core clock speed (it is a part of the 128KB L1 cache on the Athlon).The CC Winstone 2K test is very memory dependent since the test simulates real world usage by multitasking a number of content creation programs. Unfortunately this penalizes the VIA Apollo Pro 133A test bed that isn't able to keep up with the more expensive RDRAM setup in terms of memory performance.

SYSMARK2000 Performance 


The situation under SYSMark 2000 is far from as close as it was under CC Winstone 2000. The SYSMark 2000 tests definitely benefit more from SSE and RDRAM than the CC Winstone 2000 tests did and it shows by the huge lead the Pentium III at 1GHz assumes over the 1GHz Athlon. Even the VIA 133A setup with the 1GHz Pentium III managed to pull ahead of the Athlon at 1GHz here.

Quake III Arena Performance
 
While the Athlon at 1GHz was the fastest thing we had ever tested two days ago, the Pentium III at 1GHz comes in and takes performance to a new level. At 640 x 480 we are able to limit the effects of the graphics subsystem on the tests and we can truly stress the CPU, in this case we have the Pentium III 1GHz pulling ahead of the rest of the competition, including the 1GHz Athlon.Both of these options, the 1GHz Athlon and the 1GHz Pentium III are most likely too expensive for the likes of most gamers, but keep in mind that in a few months these 1GHz chips will become the affordable bunch.

 

Here's something very interesting. The Pentium III at 1GHz on the more expensive i820 + RDRAM setup falls slightly below the same CPU on a 133A board. This advantage is next to nonexistent in real world gameplay but the fact that it is here is what raises interest. What's even more interesting is that the performance advantage disappears when you use drivers other than the newest 3.76 detonator drivers that apparently improve performance on the KX133 that uses the same AGP core as the VIA 133A.

UnrealTournament Performance
 

UnrealTournament also favors the Pentium III 1GHz, even on the more affordable VIA setup, to the 1GHz Athlon. The performance difference here obviously isn't tremendous, but do remember that the UnrealTournament engine has a few limitations that are kicking in here that unintentionally make sure the performance difference between two processors is kept to a minimum, or two video cards for that matter.

 

Expendable Performance
 

Although it seemed like the 1GHz Athlon could pull ahead here, the 1GHz setup from Intel took the crown in Expendable. If you toss out the RDRAM setup because of its cost, then the 133A setup does take a back seat to the Athlon/KX133.
 

 SPECviewperf
The Standard Performance Evaluation Corporation, commonly known as SPEC, managed to come up with a synthetic benchmark with real world implications. By running specific "viewsets" SPECviewperf can simulate performance under various applications. To be more accurate, according to SPEC, "A viewset is a group of individual runs of SPECviewperf that attempt to characterize the graphics rendering portion of an ISV's application." While this method is by no means capable of identifying the performance of a card in all situations, it does help to indicate the strengths and weaknesses of a particular setup.SPECviewperf 6.1.1 currently features five viewsets: the Advanced Visualizer, the DesignReview, the Data Explorer, the Lightscape and the ProCDRS-02 viewset. Before each benchmark set we've provided SPEC's own description of that particular viewset so you can better understand what that particular viewset is measuring, performance-wise.

 Each viewset is divided into a number of tests, ranging from 4 to 10 in quantity. These tests each stress a different performance element in the particular application that viewset is attempting to simulate. Since all applications focus on some features more than others, each one of these tests is weighted meaning that each test affects the final score differently, some more than others.All results are reported in frames per second, so the higher the value, the better the performance is. The last result given for each of the viewsets is the WGM or Weighted Geometric Mean. This value is, as the name implies, the Weighted Geometric Mean of all of the test scores. The formula used to calculate the WGM is as follows:

With n being the number of tests in a viewset and w being the weight of each test expressed as a number between 0.0 and 1.0.If you'd like to know more about why a Weighted Geometric Mean is used, SPEC has an excellent article detailing just why,  here .We ran the SPECviewperf 6.1.1 package under NT for a high-end workstation performance comparison. A noteworthy change has been made since the last Athlon CPU review, we have started using a DDR GeForce in the high-end tests after  discovering  that in spite of the GeForce's hardware T&L engine the card uses virtually all of the host CPU during the SPECviewperf tests thus putting a great strain on the CPU.We also received an updated set of drivers (v3.76) for Windows NT that fully take advantage of the KX133 chipset which help to produce some very interesting benchmark results. An interesting thing to note is that because the VIA Apollo Pro 133A (VIA 133A) chipset uses the same AGP core as the KX133 chipset, the 3.76 drivers actually improved performance on the 133A platform as well.
 
Advanced Visualizer (AWadvs-03) Viewset
Taken from www.spec.org/gpc/opc.static/awadvs.htm

Advanced Visualizer from Alias/Wavefront is an integrated workstation-based 3D animation system that offers a comprehensive set of tools for 3D modeling, animation, rendering, image composition, and video output. All operations within Advanced Visualizer are performed in immediate mode with double buffered windows. There are four basic modes of operation within Advanced Visualizer:

  •  
  • 55% material shading (textured, z-buffered, backface-culled, 2 local lights)
    • 95% perspective, 80% trilinear mipmapped, modulated (41.8%)
    • 95% perspective, 20% nearest, modulated (10.45%)
    • 5% ortho, 80% trilinear mipmapped, modulated (2.2%)
    • 5% ortho, 20% nearest, modulated (.55%)
  • 30% wireframe (no z-buffering, no lighting)
    • 95% perspective (28.5%)
    • 5% ortho (1.5%)
  • 10% smooth shading (z-buffered, backface-culled, 2 local lights)
    • 95% perspective (9.5%)
    • 5% ortho (.5%)
  • 5% flat shading (z-buffered, backface-culled, 2 local lights)
    • 95% perspective (4.75%)
    • 5% ortho (.25%)


 

The Awadvs viewset is definitely showing limitations of the video card as the performance barely changes from the Pentium III 600/133 to the 1000/133. The VIA 133A setup does edge out the i820 setup because of the new 3.76 drivers with their optimizations for the AGP core used in the 133A chipset.

DesignReview (DRV-06) Viewset
Taken from www.spec.org/gpc/opc.static/drv.htm

DesignReview is a 3D computer model review package specifically tailored for plant design models consisting of piping, equipment and structural elements such as I-beams, HVAC ducting, and electrical raceways. It allows flexible viewing and manipulation of the model for helping the design team visually track progress, identify interference, locate components, and facilitate project approvals by presenting clear presentations that technical and non-technical audiences can understand. There are 6 tests specified by the viewset that represent the most common operations performed by DesignReview.
 

 

While the Pentium III may come out on top in the content creation tests, the Athlon continues to take the crown in the professional level OpenGL arena. The DRV-06 viewset is definitely not as limited by the graphics subsystem as the previous test, and because of that we see a clear difference between the various processors. The 1GHz Pentium III does come in third place, just under the Athlon 950. The RDRAM test platform drops even further below that point to below the Athlon 800. Once again, it seems as if the 3.76 drivers are doing some helpful optimizations for chipsets that use VIA's AGP 4X core.Data Explorer (DX-05) ViewsetTaken from: www.spec.org/gpc/opc.static/dx.htm

The IBM Visualization Data Explorer (DX) is a general-purpose software package for scientific data visualization and analysis. It employs a data-flow driven client-server execution model and is currently available on Unix workstations from Silicon Graphics, IBM, Sun, Hewlett-Packard and Digital Equipment. The OpenGL port of Data Explorer was completed with the recent release of DX 2.1.The tests visualize a set of particle traces through a vector flow field. The width of each tube represents the magnitude of the velocity vector at that location. Data such as this might result from simulations of fluid flow through a constriction. The object represented contains about 1,000 triangle meshes containing approximately 100 vertices each. This is a medium-sized data set for DX.
 


 
Performance here is much worse for the 1GHz Pentium III as it falls below the Athlon at 800MHz on the KX133. For all of you Athlon owners that are having fun with these numbers, remember that the minute you benchmark an Athlon on an AMD 750 board, the performance suddenly drops from being best of the best to the absolute worst.NVIDIA's 3.76 drivers obviously do quite a bit of the KX133 and the 133A, isn't it amazing what 6MB of well written drivers can do?
 
Lightscape (Light-03) Viewset
Taken from: www.spec.org/gpc/opc.static/light.htm

The Lightscape Visualization System from Discreet Logic represents a new generation of computer graphics technology that combines proprietary radiosity algorithms with a physically based lighting interface.There are four tests specified by the viewset that represent the most common operations performed by the Lightscape Visualization System

 

Nothing too surprising here, once again the Pentium III falls behind the Athlon, this time by about 15% at 1GHz.

ProCDRS-02 Viewset
Taken from: www.spec.org/gpc/opc.static/procdrs.htm

The ProCDRS-02 viewset is a complete update of the CDRS-03 viewset. It is intended to model the graphics performance of Parametric Technology Corporation's CDRS industrial design software.For more information on CDRS, see www.ptc.com/icem/products/cdrs/cdrs.htmThe viewset consists of ten tests, each of which represents a different mode of operation within CDRS. Two of the tests use a wireframe model, and the other tests use a shaded model. Each test returns a result in frames per second, and a composite score is calculated as a weighted geometric mean of the individual test results. The tests are weighted to represent the typical proportion of time a user would spend in each mode.


All tests run in display list mode. The wireframe tests use anti-aliased lines, since these are the default in CDRS. The shaded tests use one infinite light and two-sided lighting. The texture is a 512 by 512 pixel 24-bit color image.

 

For the final SPECviewperf test using the ProCDRS-02 viewset, the RDRAM setup actually pulls slightly ahead of the 133A platform to place the 1GHz Pentium III just below the performance of the Athlon 900.

Conclusion
For content creation applications and games (in most cases), the 1GHz Pentium III is faster than the 1GHz Athlon. Whether this is because specific SSE optimizations or because of the full speed on-die L2 cache of the Pentium III depends on the particular game and/or application, but for the most part the Pentium III will come out on top.Things change once you enter the professional graphics arena, where the Athlon completely dominates provided that it is paired up with a KX133 motherboard.
 For most users, neither of these CPUs is worth the time of day simply because of their cost, whereas overclocking a slower chip to speeds close to 1GHz would make much more sense. For the Pentium III, reaching 1GHz via normal overclocking methods may be a bit difficult, but it has definitely been done before with the Athlon.In terms of motherboard platforms, the VIA Apollo Pro 133A is now, hands down, the best platform for the Pentium III. While the BX is still the most intelligent for those users that have motherboards that support the new CPUs and don't mind the fact that they run at a 100MHz FSB, for new system builders, the 133A makes the most sense, especially since there are motherboards out there based on the 133A that are pretty solid.For most of you all that will be building systems rather than buying them, the fact that the 1GHz Pentium IIIs won't be available until Q3 is a big downside, potentially a reason to go with the Athlon if you feel that you must have that sort of power soon.

 How it Rates
 

Martin,
Hardware Hackers Team
Attachments:
Last edit: 7 months 21 hours ago by Martin.

Please Log in or Create an account to join the conversation.

Time to create page: 0.338 seconds

About Hardware Hackers

Hardware Hackers, Modifying For The Greater Good! Follow us while we take you on a journey of custom computer modifying mayhem. We are available on Facebook & YouTube see our links below.

Why not join in the fun at our community forum? Visit forum or Create an account.

Subscribe For Updates

Subscribe to get our latest updates straight to your inbox and stay updated on the latest developments from the HwH Team!