Northwood eventually outclassed Athlon XP. Somewhere up around 2.8-3.0 GHz. I think that Northwood and Athlon XP were usually interchangeable otherwise, especially when you got hyperthreading because that improved multitasking noticeably. They both had their strengths. Of course, AMD was the usual "value underdog".
http://techreport.com/articles.x/3289/11

Prescott was more compared to Athlon 64 than Athlon XP. I see Prescott as similar to Barton in that the architecture was at its limits and needed to be replaced.

swaaye wrote:

You know in retrospect for Slot A Athlon vs. Coppermine, I think the enthusiast world was way too thrilled by Athlon. You got a much hotter CPU with really shitty motherboards and performance that was essentially the same or intangibly better.

Tbird was lucky that Willamette sucked hard and was hot too. 😀 Although once Willamette got to 2 GHz I think Athlon had been matched overall. And Northwood is definitely a match for Athlon XP.

But Athlon was a giant improvement over K6 junk, aside from the still awful motherboards. I suppose most of the excitement came from that aspect. Intel actually had competition in every application.

How did the Duron generally compare with Coppermine and Tualatin Celerons? On paper it looks like it gets stomped.

Back in the day they were pretty highly regarded, the Spitfire durons were supposedly not far behind the Athlons in performance. The later Morgan core durons had SSE which helped level the playing field with the P3. Especially if you used DDR RAM with them, which the P3 couldn't take advantage of. I've actually got a magazine here that compared a 1.3GHz Duron to a 1.3GHz Tualatin and the duron was slightly faster. However the P3 had SDRAM and the Duron had DDR.

Duron was comparable to P3 Coppermine (in some cases even better).
On paper it looks weak because of its 64KB L2 cache, but in contrast to P6 microarchitecture Duron cache works in exclusive mode which means that data in L1 and L2 cache are not duplicated.

swaaye wrote:

You know in retrospect for Slot A Athlon vs. Coppermine, I think the enthusiast world was way too thrilled by Athlon. You got a much hotter CPU with really shitty motherboards and performance that was essentially the same or intangibly better.

Tbird was lucky that Willamette sucked hard and was hot too. 😀 Although once Willamette got to 2 GHz I think Athlon had been matched overall. And Northwood is definitely a match for Athlon XP.

But Athlon was a giant improvement over K6 junk, aside from the still awful motherboards. I suppose most of the excitement came from that aspect. Intel actually had competition in every application.

A friend of mine used to have an Athlon 500. It was pretty solid, and definitely better then my P2-350!
And even if Northwood can match Athlon XP in real world performance, it's doing it much more inefficiently.

Socket 423 is a strange thing. I kinda compare the s423 to s478 like I compare Socket 4 to Socket 5/7.
As to why they released a socket that was planned to be replaced 6 months later (this was known even before s423 boards were for sale!), perhaps they wanted to bring Netburst in the market ASAP and needed more time to get micro-PGA in production?

Edit:Btw, you're right about 1 thing:When the P3 was Intel's flagship, Athlon was the more inefficient one if you take heat into account 😉
And no SSE

Holy crap! Look at this

http://www.spodesabode.com/archive/content/article/p3duron/2

A Duron 800 and PIII (PIII, NOT Celeron!) 700 both OC'ed to 933 mhz and the Duron beats the P3 in everything but 3DMark and the difference in 3DMark is small enough where it probably wouldn't even be noticeable in games. The Duron may even be faster in some games because synthetic benchmarks don't always tell the whole truth about a system's gaming capabilities. This was at a time when the PIII tested cost twice as much as the Duron. That effectively gives the Duron twice the bang for the buck!

sliderider wrote:

Holy crap! Look at this

http://www.spodesabode.com/archive/content/article/p3duron/2

A Duron 800 and PIII (PIII, NOT Celeron!) 700 both OC'ed to 933 mhz and the Duron beats the P3 in everything but 3DMark and the difference in 3DMark is small enough where it probably wouldn't even be noticeable in games. The Duron may even be faster in some games because synthetic benchmarks don't always tell the whole truth about a system's gaming capabilities. This was at a time when the PIII tested cost twice as much as the Duron. That effectively gives the Duron twice the bang for the buck!

Yep, clock for clock the Athlons were faster then Coppermine, and on top of that, they were winning the Mhz race also.
Intel was shitting its pants back then! 😁

What Duron was good at was beating the Celeron. Those Coppermine-based Celerons were nasty chips because of the cache reduction and the 66 MHz bus nonsense. Even at 100 MHz bus they still had a hard time because the Intel chips were more sensitive to small L2 caches.

Duron just happened to have enough cache to perform rather well whereas Celeron was underneath the happy threshold in quite a few applications. Celly had to go to RAM too much.

Tetrium wrote:

And even if Northwood can match Athlon XP in real world performance, it's doing it much more inefficiently.

It's not that inefficient though. The power consumption of Athlon XP and most Northwood P4s are similar. A clock speed comparison isn't meaningful because Athlon's architecture can't clock as nearly as high as P4 and P4 isn't designed to perform at lower clocks.

Power consumptions:
http://en.wikipedia.org/wiki/List_of_AMD_Athl … microprocessors
http://en.wikipedia.org/wiki/List_of_Intel_Pe … 8130.C2.A0nm.29

Also notice how the Athlon XP 3200+ and the P4 Northwood 3.0 GHz have a similar jump in power consumption. Both chips were nearing the edge.

Here's a link for Tualatin celeron 1300MHz vs Duron (morgan) 1200MHz. The Celeron is still bested by the Duron for mostly "Gaming". While a few other benchmarks show the Tualeron being pretty decent and still can't be touched for TDP. That Duron runs 50watts vs. Celerons 18watt!

http://www.tomshardware.com/reviews/intel,406.html

I still love the PIII-S 1.4GHz, it holds its own against the thunderbird 1400 with almost half the wattage. Good enough for Windows 98SE on most things at this point.

About the burnout, did you have the BIOS thermal protect set to enabled? Without a HS I'm sure it would suffer either way. but I've noticed a few of my intel boards don't even have thermal protection enabled, like my Soyo TISU board.

Yeah the Tualatin Celeron is pretty decent because it's mostly like a Coppermine P3 because it has 256K L2. It also has some Tualatin core tweaks as well. Tualatin was a great chip.

If you also look at Pentium M vs. Athlon 64 you can see the performance parity with the Pentium M having a huge power consumption advantage. Pentium M is just a greatly improved Tualatin. It was expensive to build a nice desktop Pentium M system, but there were excellent gaming notebooks with them.

Yes, I always liked the Coppermine/Tualatin line, but as soon as CPU computational power exceeded what Tualatin could deliver, I'm go Athlon (XP/64).
I'd love to have one of those post-netburst boards in a rig of mine, but alas, can spend every dime only once, right? 😀

Edit:And I think Intel changed from using Max Power Dissipation in Tualatin to TPD (which is always LOWER) in Netburst as a way to camouflage it's true power dissipation.
Netburst had the tendency to underclock itself when it got to a certain temperature, so when it's heatsink gets too ineffective at cooling it (dust is one reason) it would simply start performing worse.
Intel did everything it could to "talk straight that what is bend" when it's flagship was netburst.

But anyway, after netburst's ultimate failure, Intel went back to P6 architecture (after firing a lot of people on their design team, LoL!).

DonutKing wrote:

retro games 100 wrote:

I have found a 2 minute youtube video here, containing relevant information to this discussion.

This is the same video I linked to above 🤣

Sorry about that! 😦 That was an uncanny coincidence.

It's too bad those early Athlons did not have the overheat protection that the Intel chips did.

Apologies if it got mentioned before, but one chief advantage of the Athlon/Duron chips was their huge L1 cache. 4 times as much as a P3 of any kind.

But honestly, could I build a retro system around a Tualatin with open multi, I'd much rather do that than go Barton. But there is no choice it seems.

I will see if my "hot" Athlon retro crate is of any advantage - if not, I'll gladly fall back on my BX machine.

I just need to make sure not to repeat RM100s experience.

Tetrium wrote:

Netburst had the tendency to underclock itself when it got to a certain temperature, so when it's heatsink gets too ineffective at cooling it (dust is one reason) it would simply start performing worse.

My parents have a nearly silent 2.66 Northwood and I have one at work as a web kiosk. They are both Dell boxes and are cooled by a plain aluminum heatsink and a quiet ducted 92mm case outlet fan. The cases even have crap airflow as I found out when I tried putting a Prescott in one.

Prescott is the furnace. Before that core, P4 really was not that hot. But Prescott is like a quad core without the excitement of actual performance.

ux-3 wrote:

Apologies if it got mentioned before, but one chief advantage of the Athlon/Duron chips was there huge L1 cache. 4 times as much as a P3 of any kind.

It's just AMD's take on caching. Their L1 may be huge but it's also slower and less associative than Intel designs so it's less effective. It was a huge benefit for the Duron though because losing some L2 cache wasn't as big of a deal.

BTW, Sandy Bridge is doing away with AMD's decade old 128K L1 design.

ux-3 wrote:

Apologies if it got mentioned before, but one chief advantage of the Athlon/Duron chips was there huge L1 cache. 4 times as much […]
Show full quote

Apologies if it got mentioned before, but one chief advantage of the Athlon/Duron chips was there huge L1 cache. 4 times as much as a P3 of any kind.

But honestly, could I build a retro system around a Tualatin with open multi, I'd much rather do that than go Barton. But there is no choice it seems.

I will see if my "hot" Athlon retro crate is of any advantage - if not, I'll gladly fall back on my BX machine.

I just need to make sure not to repeat RM100s experience.

There is more than L1 cache at work. I remember reading about how the Athlon is more or less a copy of the P3 with about 3 noteable improvements. If I find it again I'll post it up. Even the Core cpu's use alot of the PIII techologies for the basis of its build and finally killing off netburst.

If you got the money to waste you could find a Pentium-M adapter for a S478 Desktop motherboard or the rare Asrock motherboards that would support the Pentium-M. Then run the 2.13 or whatever was fastest of the time.

I suppose you could define AMD K7 as a beefed up P3 because it does the same stuff just in wider quantities. 😉

Good articles:

Pentium 4: Round 1 - Intel blows the lead
Pentium 4: Round 2 - The Ass Kicking Continues
Pentium 4: Round 3 - AMD hangs on to lead
http://www.emulators.com/docs/pentium_1.htm

7th Generation CPU Comparisons
http://www.azillionmonkeys.com/qed/cpujihad.shtml

6th Generation CPU Comparisons
http://www.azillionmonkeys.com/qed/cpuwar.html

PowerPC on Apple: An Architectural History, Part I
http://arstechnica.com/old/content/2004/08/ppc-1.ars
http://arstechnica.com/old/content/2004/10/ppc-2.ars

😁

The K5 is pretty interesting too but it's too old to find articles about easily. It was basically the AMD 29000 RISC architecture with x86 decoding. Similar to the Pentium Pro but not as fast in the end unfortunately because they didn't have the skill to pull it off. It kicks around the Pentium though at the same clock speed.

RogueTrip2012 wrote:

If you got the money to waste you could find a Pentium-M adapter for a S478 Desktop motherboard or the rare Asrock motherboards that would support the Pentium-M. Then run the 2.13 or whatever was fastest of the time.

No ISA - no go.

swaaye wrote:

It kicks around the Pentium though at the same clock speed.

Even in FPU performance?

Tetrium wrote:

swaaye wrote:

It kicks around the Pentium though at the same clock speed.

Even in FPU performance?

Whenever I compared FPU, Intel won. It was that simple.