Main period AMD leadership was from K6-3 (or -2? from 1998) to Athlon FX (up to Core/Core 2 architecture, released at 2006).

What is a 'counterpart'? A CPU in the same priceclass, or absolute performance regardless of price?
In the former case, I think AMD nearly always delivered better price/performance. In the latter case, only Athlons/Athlon64s were actually faster than anything Intel offered. The rest of the time, AMD mainly competed with Intel's older and/or lower spec CPUs.

DURON "SPITFIRE" was the clear winner over Celeron "coppermine", cheaper too. Not high end CPU's of course, but what you'd find in many,many home built gaming rigs.
I think the Athlon "Thunderbirds" were faster than PIII "Coppermine" & early P4 "Willamette", they were actually available to buy too, unlike the coppermine CPU's that were barely available for a good while outside OEM boxes.

BSA Starfire wrote:

I think the Athlon "Thunderbirds" were faster than PIII "Coppermine" & early P4 "Willamette", they were actually available to buy too, unlike the coppermine CPU's that were barely available for a good while outside OEM boxes.

They are. FPU performance at least, on a 1333MHz athlon beats a 1.4GHz tualatin hands down - both in synthetics and games. DK2 alone (@1600x1200, 32bit, max detail) runs fluently on said 1333MHz Athlon / Abit KT133 / GF4 Ti4200, while my 1.4GHz tualatin / Asus ti4600 struggles badly at the same resolution, getting massive framerate drops when zooming out and when lots of creatures are on screen. Same for black and white 1. This game performs rather poorly on both systems, but is noticeably faster on the athlon.

On-topic, the Athlon and Athlon XP were faster clock per clock then their intel counterparts. The 3200+ running at 2.2GHz equals a 3GHz Prescott in most apps and games, although the P4, especially the Prescott and Gallatin have very good FPU performance and do great in games. I remember back in the day, an Athlon XP 1700+ @ 1.5GHz beat the crap out of the 1.7GHz Willamette. even after Northwoods came out, the entry level Athlons were still faster then equivalent P4 chips. The gap only closes on high end chips, like the aforementioned 3200+ and 3GHz P4 Prescott, with the 3.2GHz Prescott equaling the 3200+ Barton and beating in some apps/games, due to greatly increased L2 cache size (from 256kb on the Willamette to 1MB on the Prescott and 2MB on the Gallatin) as well as greatly increased FSB and clock speeds. It seems Netburst loves lots of L2 cache, high FSB and high clock speeds, and needs these to perform well.

I've recently discovered that 478 machines do have a great advantage over socket A when it comes to top-of-the-line models - overclokability. While a 1.5GHz 1700+ can sometimes overclock to 1833MHz and a 2500+ Barton will often overclock to 3200+ levels (2.2GHz from 1833MHz by setting the FSB from 166 to 200), so will equivalent intel chips. A 2.4GHz P4 will most often comfortably run at 2998MHz, and require less voltage to do so (most 2500+ chips need 1.65 to 1.7v to run at 2.2GHz stable, while the 2400Mhz northwood will do so @ 1.52v most of the time (from a stock 1.45v). It gets even worse on high performance chips. A 3200+ will run at 2.4, 2.5GHz tops, and this on unlocked chips that allow changing the multiplier, or on modded chips. At 2.5GHz a 3200+ will want 1.75-1.8v witch is a lot of juice. Your garden variety prescott will OC from 3000 to 3660MHz (240-244MHz FSB) and some "golden chips" will go even higher on a good motherboard (4GHz on a Abit IC7 @ 1.49v). On the other hand, they will make A LOT of heat and draw insane amounts of power. For example, said 3200+ running @ 2.5GHz / 1.698v will top out at 72C on a regular arctic cooling copper lite, while the prescott @ 4GHz / 1.49v will go up over 80C, even tough I was using a huge tuniq tower 120.

Later AMD launched the Athlon64 witch was clearly superior over intel's offerings at the time. AMD had the fastest chips until intel released the Centrino and Core 2 Duo series.

Ryzen vs Kabyake, very likely Threadripper vs KabylakeX and again very likely EPYC vs Xeon

kanecvr wrote:

although the P4, especially the Prescott and Gallatin have very good FPU performance and do great in games.

The FPU (legacy x87) is actually very poor on a P4. It's SSE2 where the P4 shines, and most games and special software like offline renderers, video encoders/decoders etc are heavily optimized for SSE2.

AMD often priced lower end chips quite aggressively so you could get performance for little cash but in general was normally playing catch-up in absolute performance . The whole PR system was an embarrassment though as AMD tended to be very optimistic with these numbers. The early K7 range was probably one of there most competitive high end CPU AMD had done and genuinely a very powerful chip. AMD did drag out the K7 family for far to long. The P4 2.4c was substantially better at many pro apps and could do dramatically more SETI work units a day than even a top of the line Athlon XP 3200+ . The early 64bit chips again saw AMD back at the high end of things with the very impressive K8. The later FX chips with shared FPU always struggled. These new Zen chips are quite impressive too fingers crossed AMD will launch some higher clocked models soon .

Bobolaf wrote:

These new Zen chips are quite impressive too.

They're not quite a K7 or K8 though.

Very true both the K7 and K8 were something really impressive when they came out. I can only imagine the panic at Intel when they realise AMD had actually bettered them.

In the general AMD dominated the CPU game Starting with AMD Athlon and it ended when Core 2 Duo.

Bobolaf wrote:

Very true both the K7 and K8 were something really impressive when they came out. I can only imagine the panic at Intel when they realise AMD had actually bettered them.

I don't think K7 was when Intel realized their mistake. K7 was only marginally better than the PIII (in fact, it is mostly a clone of that architecture, with a few tweaks and improvements).
Intel must have known before then that their P4 wasn't the leap forward from PIII that it was supposed to be, and that they probably would have gotten better results, at least initially, by just doing another refinement of the PIII.

It's quite ironic then that AMD's first completely new architecture since that K7 was Bulldozer, which was pretty much P4 all over again: the architecture was a failure, and they would have been better off sticking to their previous architecture.

Scali wrote:

the architecture was a failure, and they would have been better off sticking to their previous architecture.

The same goes for Netburst, and in the end Intel actually did exactly that.
Btw, Athlon wasn't some cheap clone of Coppermine/P6 like you make it out to be.
Thunderbird was perhaps marginally better, but it ran WAY hotter! But at least it could be scaled, which is something Coppermine had lots of trouble with when trying to surpass 1GHz (and Athlon could make more effective use of DDR compared to Pentium 3).

Scali wrote:

I don't think K7 was when Intel realized their mistake. K7 was only marginally better than the PIII (in fact, it is mostly a clone of that architecture, with a few tweaks and improvements).

Considering where much of the talent came from, I think i there was more replication of DEC Alpha technology, starting from the EV6 bus and through the operating units.

Bobolaf wrote:

Very true both the K7 and K8 were something really impressive when they came out. I can only imagine the panic at Intel when they realise AMD had actually bettered them.

Very true. Netburst was actually a response to Athlon as they could make it clock higher. This was what prompted Intel to well...basically they made their next CPU more stupid so it could be clocked higher "because more MHz/GHz must mean it is faster!!!11" and 'we' all fell for it!
They had to basically compensate for Netburt's stupidity by adding more cashes, which made their CPUs run even hotter and more difficult to scale, which they "solved" by making the next Pentium 4 even more stupid (increase the length of the pipelines) and compensate for this by adding SSE3 and even more cash...which made Preshott hotter again, it was one big mess Intel gotten itself into...along with their other blunder, RAMBUS.

Now Intel seems to be on their nerves again, with Ryzen. Suddenly we get i9 🤣, it's not as if Intel was incapable of doing so earlier...they simply had no reason to innovate and kept milking their cow.

Bulldozer was AMDs attempt to add HT, but the whole idea was flawed. But at least they got it right this time.

gdjacobs wrote:

Scali wrote:

I don't think K7 was when Intel realized their mistake. K7 was only marginally better than the PIII (in fact, it is mostly a clone of that architecture, with a few tweaks and improvements).

Considering where much of the talent came from, I think i there was more replication of DEC Alpha technology, starting from the EV6 bus and through the operating units.

Correct. This is where the whole DDR-thingy of Athlon was based on, which is something Coppermine could not do.

gdjacobs wrote:

Considering where much of the talent came from, I think i there was more replication of DEC Alpha technology, starting from the EV6 bus and through the operating units.

No.
It's talking DEC Alpha's operating units and putting then in a P6-like x86 microarchitecture, on an EV6-like bus. So it's like building an x86 on top of a RISC backend, which is basically what P6 also is. They also did it in virtually the same way (roughly the same configuration of pipeline in number of stages, number and type of different processing units etc... main difference is 3 ALUs instead of 2 ALUs, and the FPU being 3-way pipelined, which is where they got their slight gains over the P6, along with the faster bus).

Scali wrote:

gdjacobs wrote:

Considering where much of the talent came from, I think i there was more replication of DEC Alpha technology, starting from the EV6 bus and through the operating units.

No.
It's talking DEC Alpha's operating units and putting then in a P6-like x86 microarchitecture, on an EV6-like bus. So it's like building an x86 on top of a RISC backend, which is basically what P6 also is. They also did it in virtually the same way (roughly the same configuration of pipeline in number of stages, number and type of different processing units etc... main difference is 3 ALUs instead of 2 ALUs, and the FPU being 3-way pipelined, which is where they got their slight gains over the P6, along with the faster bus).

And Intel stole this idea from AMD to make their Netburst chip quad-pumped.
Lets just agree that AMD and Intel steal stuff and ideas from each other 😀
We all know this is the cold hard truth 😀
The only difference between AMD and Intel is that Intel has more money and more lawyers 😵

But anyway, I'm still wondering what was meant with the question kwyjibo asked. My guess is it's performance?
That question would be fun to answer 😀
Btw kwyjibo, say "Hi!" to Frank from me 😉 😜

Cyberdyne wrote:

In the general AMD dominated the CPU game Starting with AMD Athlon and it ended when Core 2 Duo.

Frankly, I wouldn't call it "dominating", more that AMD were on equal terms with Intel while Intel had more money to keep up with Athlon with this inferior Netburst architecture of theirs.
It all ended when Intel finally decided to come back to their senses and got the desktop-P6-idea out of the freezer (which they were at least not stupid enough to also had abandoned for laptops) and blow away K8 with their Core2.
After that, Intel created i3/i5/i7 and then gotten sloppy again and were lucky AMD did a Netburst on themselves when they created Bulldozer 😵

kwyjibo wrote:

First, I do not want to start a flame war 😀

What I want to know is, in order to be able to build some period correct PCs (gaming or workstation), and I am getting very confused about benchmarks, what AMD high-end processors were considered to be on pair or surpass their Intel high-end counterparts? Athlon 1 Ghz? First Opteron? First Athlon 64?

Athlon 1GHz is roughly equivalent to Pentium 3 1GHz
I'n not sure about Opteron, but my guess is that it's roughly similar in performance to Athlon 64 of the same clock frequency.
Athlon 64 @2.2GHz is roughly equal to Netburst 3.2GHz (A64 is probably a bit faster).
Athlon 64 was about 25% faster compared to Barton clock for clock. Barton 2.2GHz was maybe a little bit slower compared to a 3.2GHz Netburst chip.

I'm not sure if this is what you meant, please correct me if I'm wrong.

One thing most people miss that that there is more to a cpu then raw speed and performance.
lower power draw is a big thing in mobile systems, lower cost is a big deal in budget system too. It is not always about speed, the Duron was a very well remarked cpu in the day do in part to its low cost and overclocking ability.

And I'm amazed that no one said anything about the Am5x86.