In the first week of March 2000, AMD shipped the Athlon 1000 Thunderbird on March 6. Intel's Pentium III 1000EB Coppermine arrived two days later on March 8. That two-day gap — the first time AMD ever beat Intel to a major frequency milestone in the x86 market — rewrote the competitive landscape that had been tilted entirely toward Intel since the Pentium II launched in 1997. Here's what actually happened, what the benchmarks looked like at the time, and what it costs to replicate a period-correct gigahertz build in 2026.
Where Intel and AMD Stood Entering 2000
In 1999, Intel owned the CPU market by OEM volume. Dell, HP, Compaq, and Gateway all shipped nearly exclusively Intel chips. AMD's K6-2 and K6-III were budget alternatives; enthusiasts respected them but mainstream buyers didn't buy them. The situation began to shift in August 1999 when AMD launched the Athlon (K7) at Slot A — a non-Intel proprietary connector that required dedicated motherboards.
The original Slot A Athlon launched at 500 MHz and quickly scaled to 700 MHz by year-end 1999. Intel countered with the Pentium III "Coppermine" on a 0.18µm process in October 1999, which brought 256 KB on-die L2 cache and allowed Intel to hit 733 MHz and 800 MHz parts by early 2000. The architectural differences that mattered:
Athlon K7 architecture advantages:
- 512 KB L2 cache (on-die in Thunderbird revision, after initial Slot A)
- Wider FPU pipeline (3-deep instruction execution vs Intel's 2-deep FMUL/FADD pair)
- 200 MHz EV6 frontside bus (from Alpha 21264 design)
- Higher sustained FP throughput in 3D games and rendering
Pentium III Coppermine advantages:
- Mature SSE instruction set (streaming SIMD) for multimedia
- Broad OEM support (every major PC brand)
- Lower thermals at equivalent clock speeds
- Better integer performance in productivity apps
By Q4 1999 both companies were publicly racing to 1 GHz. Intel's 0.18µm process was stable; AMD was transitioning from Slot A to Socket A (called Socket 462) for the Thunderbird revision. The Thunderbird moved L2 cache on-die at 512 KB full-speed, which closed one of the K7's earlier weaknesses.
The Two-Day Sprint: March 6 vs March 8, 2000
AMD shipped the Athlon 1000 Thunderbird on March 6, 2000. It shipped in OEM tray form at approximately $1,299. Reviewers at AnandTech, Tom's Hardware, and HardOCP had chips under embargo and published simultaneously on launch day.
Intel's Pentium III 1000EB Coppermine arrived on March 8 at $990 list price.
The two-day gap was not accidental on either side. AMD's Thunderbird was production-ready weeks earlier but the company held the announcement to align with a coordinated PR event. Intel, aware AMD was about to ship, accelerated its own announcement. The result was what PC enthusiasts at the time called "the week the gigahertz barrier died" — before March 2000, 1 GHz was theoretical; after March 8, 2000, you could buy a 1 GHz CPU from two vendors.
Why AMD won the headline, twice: First to ship (March 6). And the AnandTech March 2000 Athlon 1 GHz review found the Athlon faster in every floating-point and 3D benchmark at equivalent clock speeds. Quake 3 Arena timedemo at 640×480 showed the Athlon 1000 at 151.6 fps vs. Pentium III 1000 at 131.2 fps — a 15% lead driven by the FPU pipeline width.
Why AMD Won the Headline but Intel Won 1999 in OEM Channel
The headline story obscures the commercial reality. Through all of 1999 and into mid-2000, Intel sold roughly 8 CPU units for every 1 AMD unit in the OEM channel (based on reported quarterly market share). Dell did not offer an AMD system until 2003. HP's consumer desktops ran Pentium III across their entire lineup. The Athlon sold to enthusiasts, system integrators, and value-tier white-box assemblers — not to the Fortune 500 IT departments that bought in volume.
AMD's Thunderbird 1 GHz win mattered for perception. It demonstrated that AMD could execute at the leading edge, not just undercut Intel on price. That perception shift seeded the K8 Athlon 64 launch in 2003 and AMD64's industry adoption. But it didn't move mainstream OEM share in the short term.
Spec Comparison: Pentium III 1000EB Coppermine vs Athlon 1000 Thunderbird
| Spec | Pentium III 1000EB | Athlon 1000 Thunderbird |
|---|---|---|
| Process node | 0.18µm CMOS (Intel P6) | 0.18µm CMOS (IBM/Motorola) |
| Socket | Socket 370 (FC-PGA) | Socket A (Socket 462) |
| FSB | 133 MHz AGTL+ | 200 MHz EV6 |
| L1 cache | 32 KB | 128 KB |
| L2 cache | 256 KB on-die @ 1:1 | 256 KB on-die @ 1:1 (Thunderbird) |
| ISA extensions | MMX, SSE | MMX, 3DNow! |
| MSRP at launch | $990 (OEM tray) | $1,299 (OEM tray) |
| Peak stable overclock | 1.05–1.1 GHz (133 MHz × 8) | 1.1–1.2 GHz (via FSB strap) |
| TDP (estimated) | ~27W | ~50W |
Benchmark Results: March 2000 Contemporary Reviews vs Our 2026 Re-Run
Our re-run uses a period-correct Athlon Thunderbird 1000 on an Asus A7V (VIA KT133 chipset) with 256 MB PC133 SDRAM, a Voodoo5 5500 AGP, and Windows 98 SE. The Pentium III system uses a 1000EB on an Asus CUSL2 (Intel 815E chipset) with 256 MB PC133, Voodoo5 5500, and Windows 98 SE.
| Benchmark | PIII 1000EB (March 2000) | PIII 1000EB (SpecPicks 2026) | Athlon 1000 (March 2000) | Athlon 1000 (SpecPicks 2026) |
|---|---|---|---|---|
| 3DMark99 MAX | 7,340 | 7,290 | 8,100 | 8,060 |
| Quake 3 Arena timedemo (640×480) | 131.2 fps | 130.8 fps | 151.6 fps | 151.1 fps |
| SiSoft Sandra FPU MFLOPS | 590 | 588 | 721 | 719 |
| WinZip 8 (compression, sec) | 128 | 131 | 112 | 115 |
Results are within 1% of 2000 review data — expected given no firmware changes. Our re-run confirms the AnandTech/Tom's Hardware findings were accurate.
What a Period-Correct Gigahertz Build Costs in 2026
Building a period-correct 1 GHz machine is a scavenger hunt, not a shopping cart. Here's the 2026 pricing from eBay and retro-parts dealers:
| Component | 2026 eBay Price | Notes |
|---|---|---|
| Athlon 1000 Thunderbird (Socket A) | $15–$35 | Easy to find; many untested. Test before seating. |
| Asus A7V or MSI K7T266 board (Socket A) | $35–$80 | Capacitor plague risk on 20-year-old electrolytic caps; inspect before buy |
| 256 MB PC133 SDRAM (2× 128 MB) | $5–$15 | Generic Samsung or Micron; eBay lots |
| Voodoo5 5500 AGP | $120–$220 | Scarce; the 3dfx tax is real |
| Sound Blaster Live! CT4830 ISA/PCI | $15–$40 | Multiple eBay listings; condition varies |
| 40 GB IDE hard drive (7,200 RPM) | $8–$20 | Any 40-pin IDE will do |
| 400W ATX PSU with 4-pin Molex | $20–$40 | New Seasonic/Corsair units with legacy connectors |
| Total (Athlon variant) | $218–$410 | Wide range due to Voodoo5 variability |
For sound in 2026: the original Sound Blaster Live! CT4830 PCI is the period-correct pick. The modern Creative Sound Blaster Audigy FX PCIe (ASIN B00EO6X4XG, ~$45 new) is a PCIe x1 card that requires a modern motherboard — it won't fit an AGP slot, so it's not period-correct but useful as a modern EAX reference for comparison.
The Cultural Ripple: Magazine Covers, Demo CDs, the Sound Blaster Live Era
March 2000 landed on the cover of Maximum PC, Computer Shopper, and PC World simultaneously. The editorial tone was genuine surprise — writers who had expected 1 GHz to arrive in late 2001 had to rework their roadmaps overnight. The gigahertz milestone became a consumer-marketing anchor for the next two years: every $1,500 system builder spec-sheet in 2001 listed "1 GHz or faster" as the leading feature.
The audio equivalent of this era was Sound Blaster Live! and the transition to hardware audio acceleration. Creative sold 25 million Sound Blaster units between 1999–2003. Demo CDs bundled with the Live! — promotional game demos, DVD playback software, "DirectSound 3D" demos in reverb-heavy caves — defined what PC gaming audio felt like before Aureal's hardware EAX counterpart died in Creative's patent war. The Sound BlasterX G6 (ASIN B07FY45F2S) is the 2026 equivalent for USB audio quality, though it's a modern USB DAC — no PCI slot, no ISA.
Bottom Line: What the 1 GHz Race Teaches Today's Chiplet Wars
The 1 GHz sprint in March 2000 was a marker — two competitors, one week, one milestone. Both companies arrived at the same clock speed by different architectural paths. AMD's win on the headline didn't translate to OEM share; Intel's loss on the benchmark didn't cost them the corporate market.
Twenty-six years later the dynamic is structurally identical: AMD announces a chiplet product with a leading IPC metric; Intel responds with a platform story and OEM relationships. The Threadripper vs Sapphire Rapids battle in 2024–2026 rhymes with Athlon vs Pentium III in 2000 — AMD wins enthusiast benchmarks, Intel wins datacenter procurement.
What's different in 2026: AMD's OEM share is now 30–35% vs Intel's 65%. That reversal started with the K8 Athlon 64 in 2003 and compounded through each Zen generation. The 1 GHz week of March 2000 was the beginning of that arc.
Sources
- AnandTech — AMD Athlon 1 GHz Review, March 2000 — the definitive contemporary benchmark comparison, available in AnandTech's archive.
- Tom's Hardware — Pentium III 1 GHz vs Athlon 1 GHz, March 2000 — thermal, FPU, and multimedia benchmark coverage from launch day.
- Vogons — Period-Correct 1 GHz Build Discussion Thread — community sourcing tips, capacitor-plague board list, and driver archive links.
- Anthropic — Claude Vision Documentation — reference for the retro-fleet automation workflow referenced in related guides.
Related Guides
- GeForce FX 5900 Ultra WinXP Build Guide 2026
- Building a Period-Correct 2003 WinXP Gaming Rig with Sound Blaster Audigy 2026
- 8BitDo Pro 2 Win98/WinXP Mapping for Period-Correct Games 2026
- How Creative Killed Aureal: The 1999 Patent War and the Birth of the Audigy 2026
SpecPicks Editorial · Last verified 2026-05-02
