From the release of the Pentium II in 1997 through the release of the Pentium III 600 back in 1999 Intel seemed to have everything under control.  Their processors were always released on time, and upon their announcement they were immediately available for purchase both in OEM systems as well as boxed processors by end users.  Motherboard manufacturers had no problem supporting Intel's processors with motherboards since they could always depend on Intel to release CPUs that would sell and increase demand for the motherboards that would be required to run them. 

On the flip side of the coin, AMD was dominating little more than the low end processor market.  While their K6-2 and K6-III CPUs managed to find their way into quite a few systems, and in spite of AMD's introduction of their 3DNow! instructions that helped to address their problem of poor FPU performance they simply could not compete with Intel's power in terms of obtaining a large portion of the market share.  Part of the reason behind this was that the chipset support behind AMD's K6-2/K6-III platform was supplied almost exclusively by VIA and ALi, both of which had a decent amount of problems of their own (i.e. AGP compatibility), at least upon their introduction. 

Unfortunately for Intel, things began to change in September of 1999.  The date was 9/27/99 and it was intended to be the launch date of the successor to the BX chipset that had been alive for close to 1.5 years, the i820.  The i820 was intended to be launched alongside the new Pentium III B processors that day, the P3 B processors being nothing more than 133MHz FSB versions of the regular Pentium III based on the Katmai core (512KB 1/2 speed L2 cache).  The Pentium III B made its way out the door, but the i820 was no where to be seen. 

To borrow a few lines from our review of the i820 chipset that launched that day:

"It’s the chipset we’ve all been told to wait for, and it’s finally here, the elusive "Camino" chipset has finally been released as the i820 on this day, the 27^th of September in 1999.

Actually, no, it hasn’t. Due to a bug just recently discovered in the chipset, the i820 has been, as you’ve probably already heard, delayed until further notice. "

At that point, Intel had already given up the desktop performance throne to AMD and the fact that their next-generation chipset platform was not ready didn't help the situation at all.

A month later Intel released the Pentium III E and EB chips based on the Coppermine core (256KB full speed L2 cache), and it helped to boost performance for the Pentium III platform in comparison to the Athlon.

Over two months ago Intel's Pentium III finally hit the magic 1GHz mark (1000MHz), and for the past few weeks Intel has been announcing CPUs that fall in the 800MHz – 1GHz range in order to fill the gaps that existed in their CPU line.  Today we have the final Pentium III slower than 1GHz that will be announced/released by Intel, the Pentium III 933. 

We're just going to give you a quick tour of the specs of the Pentium III 933 since it isn't any different from its predecessors other than its clock speed and respective clock multiplier before getting into the benchmark comparison.

Pentium III 933 Specifications

·        29 million transistor 0.18-micron Coppermine core

·        933MHz clock speed – 7.0x 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

The 933MHz Pentium III is not any different from the rest of the Pentium III line; it is based on the 0.18-micron Coppermine core and actually uses the same stepping (cB0) of the core as the recently announced Pentium III 1GHz processor (also indicating that the 933 should be able to hit 1GHz easily).   

Since the 933 is based on the new core revision of the Pentium III it also runs on a slightly increased core voltage setting than the 866MHz and slower processors, 1.70v instead of 1.65v.  At 1.70v the 933 is still within the operating limits of the Pentium III core so there isn't a reason to worry about the increased core voltage, it is simply a way of accommodating for the higher clock speed.  This is essentially what we do when we overclock our chips and attempt to increase their stability by boosting the core voltage, except when Intel or AMD does it they consider it to be increasing the yield on their higher clock speed parts.

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 (ASB). 

The ATC on the Pentium III is just the fancy name for the on-die 256KB L2 cache.  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

The Test

Windows 98 SE Test System
Hardware
CPU(s)	Intel Pentium III 800E Intel Pentium III 700E Intel Pentium III 600E	Intel Pentium III 1.0 GHz Intel Pentium III 933 Intel Pentium III 866 Intel Pentium III 800 Intel Pentium III 733 Intel Pentium III 667 Intel Pentium III 600EB		AMD Athlon 1GHz AMD Athlon 950 AMD Athlon 900 AMD Athlon 850 AMD Athlon 800 AMD Athlon 700 AMD Athlon 600 AMD Athlon 500
Motherboard(s)	ABIT BE6	AOpen AX6C	ASUS P3V4X	ASUS K7V-RM
Memory	128MB PC133 Corsair SDRAM	128MB PC800 Samsung RDRAM	128MB PC133 Corsair SDRAM	128MB PC133 Corsair SDRAM
Hard Drive	IBM Deskstar DPTA-372050 20.5GB 7200 RPM Ultra ATA 66
CDROM	Phillips 48X
Video Card(s)	NVIDIA GeForce 256 32MB DDR (default clock - 120/150 DDR)
Ethernet	Linksys LNE100TX 100Mbit PCI Ethernet Adapter
Software
Operating System	Windows 98 SE
Video Drivers	NVIDIA GeForce 256 - Detonator 3.76 @ 1024 x 768 x 16 @ 75Hz VIA AGP GART Drivers v3.59 was used for all VIA based boards
Benchmarking Applications
Gaming	GT Interactive Unreal Tournament 4.04 UTbench.dem idSoftware Quake III Arena demo001.dm3 Rage Software Expendable Timedemo
Productivity	BAPCo SYSMark 2000 Ziff Davis Content Creation Winstone 2000

 

Windows NT SP6.1a Test System
Hardware
CPU(s)	Intel Pentium III 800E Intel Pentium III 700E Intel Pentium III 600E	Intel Pentium III 1.0 GHz Intel Pentium III 933 Intel Pentium III 866 Intel Pentium III 800 Intel Pentium III 733 Intel Pentium III 667 Intel Pentium III 600EB		AMD Athlon 1GHz AMD Athlon 950 AMD Athlon 900 AMD Athlon 850 AMD Athlon 800 AMD Athlon 700 AMD Athlon 600 AMD Athlon 500
Motherboard(s)	ABIT BE6	AOpen AX6C	ASUS P3V4X	Gigabyte GA-7IX	ASUS K7V-RM
Memory	128MB PC133 Corsair SDRAM	128MB PC800 Samsung RDRAM	128MB PC133 Corsair SDRAM	128MB PC133 Corsair SDRAM
Hard Drive	IBM Deskstar DPTA-372050 20.5GB 7200 RPM Ultra ATA 66
CDROM	Phillips 48X
Video Card(s)	NVIDIA GeForce 256 32MB DDR (default clock - 120/150 DDR)
Ethernet	Linksys LNE100TX 100Mbit PCI Ethernet Adapter
Software
Operating System	Windows NT4 Service Pack 6.1a
Video Drivers	NVIDIA GeForce 256 - Detonator 3.76 @ 1024 x 768 x 32 @ 75Hz
Benchmarking Applications
Professional	SPECviewperf 6.1.1

The Pentium III 933 falls directly in between the 866 and 1GHz parts, obviously, but outside of the Pentium III line the 933 seems to be just as fast as the Athlon 900/950.

Contrary to popular belief, SYSMark 2000 is 3DNow! optimized in addition to being SSE optimized, so saying that SYSMark 2000 is biased against the Athlon because it features SSE optimizations but no 3DNow! optimizations isn't the best way of explaining the benchmarks.

Instead, it is correct to say that SYSMark 2000 is highly dependent on a fast L2 cache. One way we can prove this is by comparing a Pentium III 550 (Katmai) to a Pentium III 550E (Coppermine) under SYSMark 2000. Both of those CPUs run off of a 100MHz FSB, and both feature SSE instructions, the only difference being that the Pentium III 550 (Katmai) runs its 512KB of L2 cache at 275MHz (1/2 core speed) while the 550E (Coppermine) runs its smaller 256KB of L2 cache at 550MHz (full speed, on-die). The 550 (Katmai) ends up scoring a 112 here on a BX motherboard, and on the same exact test bed the 550E (Coppermine) scores a 127, an improvement of over 13% entirely due to the lower latency, higher clocked L2 cache of the 550E.

And any test that appreciates a fast L2 cache will obviously appreciate high memory performance, we already know that the VIA Apollo Pro 133A doesn't have the world's fastest memory controller, thus helping to explain the small 820 + RDRAM lead over the VIA platform.

It is for this reason that the benchmark comes out showing that a Pentium III 866 is faster than an Athlon running at 1GHz. Wait a few more weeks and watch how the Thunderbird (Athlon with on-die L2) performs under this very L2 cache dependent test suite.

In order to allow for comparison to previous CPUs, we refrained from switching to 5.xx Detonator drivers for this test, but this will be the last CPU test we use the now aging 3.xx drivers for gaming performance tests. The new 5.xx drivers, aside from enabling GeForce 2 GTS support, also improve performance by a small but noticeable (at least on graphs) for the AMD/VIA based platforms, which does change the standings a bit.

The only thing you need to draw from this chart is the relative performance of the 933, being slower than the 1GHz yet faster than the 866 (which you could guess without ever seeing a set of benchmarks) and should be somewhere in between the Athlon 900 and the 950 using the 5.xx drivers.

Here the test bed's GeForce DDR is fill rate limited which explains the relatively stagnant scores.

The situation under UnrealTournament is much like that under Quake III Arena, the 933 is performing very well but that's what you'd expect in the first place.

Fill rate limitations kick in slightly but not incredibly because of the nature of the Unreal Tournament engine, which is why the standings mimic those at 640 x 480.

Under Expendable the standings are more evenly split between the Pentium III and the Athlon, this time with the 933 weighing in between the Athlon 950 and the 1GHz monster.

We have a similar situation at 1024 x 768 since Expendable isn't really a fill rate limited game, rather a CPU/memory bound benchmark.

Ever since the Athlon was able to be paired up with the KX133 chipset it has dominated virtually the entire suite of SPECviewperf benchmarks, although with the video card being a limitation in this test, there is a very small difference between the Pentium III 600/133 and the fastest Athlon 1GHz system here.

The lead goes to the Athlon here as usual, but the Pentium III 933 is performing quite respectably as well.

In the Data Explorer test there is a much more defined performance difference between the Athlon and the Pentium III. As usual, the 933 performs slightly above the 866 and slightly under the 1GHz chip.

The Lightscape test definitely goes to the Athlon, as the Pentium III falls a noticeable distance behind it on a clock for clock basis.

Concluding the SPECviewperf tests we see the same trend continuing.

Conclusion

Both Intel and AMD have beat their product lines to death, but at least for the latter we'll be seeing a bit of a refresher with the upcoming Thunderbird/Duron releases. The Pentium III still has a few more months in it before it is dwarfed by the Willamette, in these next few months Intel will definitely have a difficult time pitting the Pentium III up against the Thunderbird unless they can bump up the clock speed to beyond 1GHz and can do so while dropping prices on the entire Pentium III line in order to remain competitive with the new Athlons.

The Pentium III 933 is actually a pretty interesting chip in that it should easily be able to make it up to 1GHz and beyond, whereas the 866 wasn't a guarantee since it was based on the old stepping of the Coppermine core (cA2). The Pentium III 933 is essentially a 1GHz part, it operates at the same core voltage, uses the same stepping of the core and should easily be able to hit that 66MHz higher clock frequency. But if you're in the overclocking mood, you'll probably want to opt for a slower Pentium III and take that up to a higher point, such as the Pentium III 600E (100MHz FSB) and push it to 800EB (133MHz FSB). The problem with overclocking something like the 933 is that you can only really successfully do so if you have FSB settings available that lie between 133MHz and 150MHz due to its fairly high clock multiplier (7.0x).

As far as platforms are concerned, we've already proven that the Intel BX chipset at 133MHz is the highest performing platform out there for use with the Pentium III, so currently there's no excuse to even bother with an Apollo Pro 133A board and there's definitely no reason to pursue an i820 board, even if it is one that supports SDRAM, especially with all of the recalls on MTH chips that have been going on recently. This takes us back to the original recommendation we made when we first reviewed the Coppermine last year, stick with the BX platform.

The Pentium IIIs are finally more available than they were a couple of months ago. As of the time of publication you can actually find Pentium III 933 chips online, although at only a few places and at a pretty hefty pricetag. The main point is that you can actually get all the Pentium IIIs (with the exception of the 1GHz parts) if you're really interested in picking one up. While that may sound like a silly thing to boast about, just a couple of months ago picking up anything faster than a 733MHz part was almost unheard of.

Then again, if you don't mind waiting, the Thunderbird is just around the corner. Provided that it gets appropriate motherboard support (which it most likely will), you can expect AMD to step forward and grab an even larger portion of the market share in the coming months before Intel can crank out their Willamette.

How it Rates