Athlon XPs 101

[Terminat.]

Note: Apologies if this has been written before, but in a recent poll (http://www.ocforums.com/showthread.php?t=350140&highlight=Poll) I found that an overwhelming majority of AMD users still use Athlon XPs, most notably of the Barton core. I hope none of you think me presumptuous for writing the “101” on Athlon XPs, but in the AMD section I find a real mention of only the AMD64s in stickies, and I apologise in advance for any minor mistakes I may make.

This is merely part one. If I receive a positive response, I’ll go on to discuss compatible motherboards, how XPs compare to Intel cpu’s and how to overclock. Plus more…

And why XPs are still a great buy for the budget pc gamer

But, if I don't get a great response, sorry for boring you.

Hope you enjoy this

Although the Athlon XP may be outdated, it is still by far the most used! This is for any user who is new to computing and owns an Athlon XP system.


Technical Terms
FSB = Front Side Bus. This is the speed at which the cpu communicates with the motherboard’s memory controller, via the Northbridge chip. FSB*Multiplier = Total CPU Speed

CPU = Central Processing Unit

North/Southbridge Chips = These are the two controllers on the motherboard. Traditionally, the northbridge controls the fast devices such as the cpu, memory, disk drives and so forth – whereas the southbridge controls the slower devices such as the mouse and keyboard.

Cache: A very small amount of memory built directly onto the cpu, allowing it to access previously completed calculations extremely quickly. There are three levels: L1, L2 and L3 – but the one most referred to is L2.

History of the Athlon XP
AMD (acronym for Advanced Micro Devices) was relatively unknown until the release of the original AMD Athlon processor – the first and last cpu to rival the Pentium 3. It was also the first processor to boast a dual-pumped FSB (more on that later.) In short, the original Athlon range was a roaring success.

But times moved on. Intel released the Pentium 4, which although hampered by its initial Willamette range, boasted an impressive technical spec. At this point, Intel was also trying to push the expensive Rambus memory in as a pc standard. Unless AMD acted, Rambus and Pentium 4 threatened the existence of the budget-cpu.
The Athlon XP 1500+ was the first of the XP range to be released. It worked with a new standard of memory that AMD pushed forward – DDRAM. This would eventually succeed in replacing the SDRAM standard, making Rambus obsolete. The difference was that Rambus was very expensive and boasted supposedly better performance, while DDRAM was considerably cheaper (fitting in line with AMD’s budget cpu.)
The XP 1500+ ran at 266Mhz FSB whereas Intel’s prototype Pentium 4 Willamette 1400 ran at 400Mhz FSB.

It is crucial to understand that AMD processors have always run at slower frequencies than their Intel counterparts, yet are capable of matching their rivals with less megahertz due to increased efficiency.
Previously AMD had labelled their processors by their actual clock speed: e.g. Athlon 1Ghz. But now they re-named their cpu’s to show the actual speed: e.g. Athlon XP 1500+ now meant that although the cpu might run at 1.2Ghz, it was the equivalent to an Intel P4 1.5Ghz, at least in the eyes of AMD.

The Athlon XP microprocessor had been born.

Cores
There are several different types of core in the Athlon XP range, listed in chronological order from oldest to newest:

Morgan: Later used for the Duron (budget series, rival to Celeron) these processors boasted a FSB of 266Mhz only. These went up to “1800+.”

Palomino: A newer processor and a replacement for the Morgan. This series of processor went all the way up to 333Mhz FSB, and extended the cpu’s L2 cache to 256k.

Thoroughbred: Again named after a racehorse, this processor was a superior version of the Palomino – and with it came increased speeds, all the way up to a “2800+” (2.09Ghz.) By this time, Quanti-Speed Architecture had been fully developed, allowing Athlon XPs to run even closer to their Intel counterparts.

Barton: The newest type of desktop core. These were identical to Thoroughbreds, except that they had double the amount of cache (512k.) This range extended all the way up to “3200+” (2.2Ghz) and also included two models running at 400Mhz FSB: the “3000+” and the “3200+” and there was also a “3000+” that ran at 333Mhz FSB.

Mobile Bartons: Abbreviated to XP-M, these processors are a rival to the P4-M range and include various technologies optimised for laptops – including the ability to reduce their clock speeds to conserve battery life. However, they quickly became known as excellent overclockers and as such were soon put to use in desktop pc’s.

 

[C_MAC]

My next chip will most likely be an XP-M. In some cases, you can save alot going the XP route vs. A64 and in other cases the prices are very similar, but In my case i'd be saving alot. They're great chips, and I wish some people woudn't look down on them as the obsolete chip. No doubt, A64's are faster at stock speeds (in some cases much faster), but that doesn't make the xp's a bad chip necessarily. Btw, it's the most used because it has been around much longer than the newer A64 chips. I was actually surprised to see the amount of A64's. Eventually, they will become the norm though.

 

[The Coolest]

That's a nice an informative thread, but the cores part could use a little improvement.
First of all Durons were all 200MHz FSB, and they never had PR ratings, the Morgan Duron topped out at 1.3GHz and had 64KB L2 cache.

The Palomino was not an improvement over the duron, it just had the full blown 256KB of L2 cache, and wasn't considered the budget line CPU. Palominos also never came with a 333MHz FSB, all of them till the last one, a 2100+ were 266FSB (133x13).
Both Palomino and Morgan were 0.18µm CPUs.

The Tbred was the first AMD CPU to be manufactured on the 0.13µm process.
The Tbred also had two stages, stepping A0 and B0.
The A0 stepping, reffered to as simply TbredA was a pretty hot, not really overclockable chip, it did a little better than the palominos but not that impressive.
The B0 stepping, the TbredB, changed all that when it was introduced, the first steppings were not too good, but ran a lot cooler, and overclocked better, very soon they were hitting 300-500MHz higher speeds than the TbredAs.

 

[Quailane]

That's a nice an informative thread, but the cores part could use a little improvement.
First of all Durons were all 200MHz FSB, and they never had PR ratings, the Morgan Duron topped out at 1.3GHz and had 64KB L2 cache.

The Palomino was not an improvement over the duron, it just had the full blown 256KB of L2 cache, and wasn't considered the budget line CPU. Palominos also never came with a 333MHz FSB, all of them till the last one, a 2100+ were 266FSB (133x13).
Both Palomino and Morgan were 0.18µm CPUs.

The Tbred was the first AMD CPU to be manufactured on the 0.13µm process.
The Tbred also had two stages, stepping A0 and B0.
The A0 stepping, reffered to as simply TbredA was a pretty hot, not really overclockable chip, it did a little better than the palominos but not that impressive.
The B0 stepping, the TbredB, changed all that when it was introduced, the first steppings were not too good, but ran a lot cooler, and overclocked better, very soon they were hitting 300-500MHz higher speeds than the TbredAs.

You are pretty much right, except that I thought the thoroughbred A's ran the same temps as the B's at the same speeds. The only difference was that the B had an extra layer in the core to stabilize it, allowing it to clock higher. The pally's were really hot, but were able to take higher temps better than the thoroughbreds.

You missed the Thorton (sp?) core, which is a barton with half the cache disabled.

It is crucial to understand that AMD processors have always run at slower frequencies than their Intel counterparts, yet are capable of matching their rivals with less megahertz due to increased efficiency.

I definately remember the AMD athlon hitting higher speeds than the p3 when they were being released.

 

[David Coleman]

Applebred Durons are 266 FSB. Tbred based core with disabled L2 cache (only 64k)

 

[c627627]

Nice read.

• Palominos never went to 333.

• T-Breds went to 2250 MHz. (The rare 2800+ T-Bred)

• There now exist 400 FSB T-Breds (found in OEM machines only, click here.)


• Athlon XP cores included Thorton besides Palomino, T-Bred A, T-Bred B and Barton.

• Morgan is Palomino with only some cache enabled. It was used in Durons only, not Athlon XPs.

etc.

 

[The Coolest]

Nice read.

• Palominos never went to 333.

• T-Breds went to 2250 MHz. (The rare 2800+ T-Bred)

• There now exist 400 FSB T-Breds (found in OEM machines only, click here.)


• Athlon XP cores included Thorton besides Palomino, T-Bred A, T-Bred B and Barton.

• Morgan is Palomino with only some cache enabled. It was used in Durons only, not Athlon XPs.

etc.

Thortons and Applebreds aren't new cores really. The Morgan physically had 64KB L2, there was no disabled cache on it, it was a different core than Palomino, exactly the same design, just less cache on the die

 

[c627627]

Your are correct. Thank you.

Funny how, there's a 'core name' Thorton & Applebred, yet there isn't a 'core name' for new x52 Opterons: http://www.c627627.com/AMD/OpteronAthlon64/
which deserve a 'core name' just as much as Venice does, don't you think?


David Coleman is also right about the 266 FSB Applebred Durons.

First of all Durons were all 200MHz FSB.

 

[The Coolest]

Yeah you're right, but I was talking about the Morgan durons... I might have not been very clear on that, I'm sorry.

 

[Splash]

Useful stickies take more than an opening line, such as XP 101. You can get there yet, with hard and useful grunt work.

For example, I would prefer this thread open with a brief history on the P3/Athlon MHZ race, why the athlon won, then the XP put the hammer down and excelled (mhz, plus prefetching, plus DDR, then the DDR race from 100-133-166-200, and beyond, and perhaps a bit of history on RAM timings importance, stretching back to SDRAM (in this old-timers!hoot! history), the chipset race with highlights on what significant offerings (Via266, A) brought and why nF2 won out over VIA, and the latest greatest nF2, mcp-400 mcp2, what dual channel brought to the table, and the best of breed mobos, MSI and Abit NF7 and others that can reside in that same spotlight. You need just a sentence for each, and a solid reference to an authoritative link to an article or messageboard thread that proves the point you make, at this point (books will be written later about such MHZ races, and the role of enthusiast communities such as ours, god willing).

You can't really do justice to an XP 101 history until you get the timeline and possibly sales of released Palemino, TbredA, TbredB, Barton. Then mult.locking, (etc) and then finally the XP-Mobile-2500-2600, seemingly best of breed, unlocked.

A touch of overclocking results from each breed, highlighting standard air overclocking results(at X(mhz) at Y(voltage) and Z(load temp), then a sample of crazy OC'ers using water cooling, and a touch of extreme OC using peltier/CO2/etc types, but since most users don't go beyond air, little sense spending more than two sentences on the extremes from each version of the XP chip.

Ultimately a satisfying conclusion would be these 3 parts (choice cpus + best chipset and 3-4 mobos + optimum RamChoices)
are the best we know, for the XP line, which seems to be at a finish, but is still, with these best parts, a hellish contender in today's power user toolbag, and still an FPU monster for distributed computing and a great choice for Grandma today.

That's my idea of a sticky.

 

[Miguita]

Traditionally, the northbridge controls the fast devices such as the cpu, memory, disk drives and so forth – whereas the southbridge controls the slower devices such as the mouse and keyboard.

Huh? The northbridge interacts with the memory and AGP. The southbridge controls sound, USB, FW, storage, etc.

 

[N3buchan3zzar]

I'm with splash on this idea. Perhaps the best thing is to put plenty of links down for those who don't know what a search button is . There is a plethora of info on these cpu's and I'm still resisting the urge to go A64 cuz mine still performs excellently.

 

[David]

My understanding of the history of the XP:

Around 1999/2000 AMD was pushing the K6-2 and K6-3 as their chips of the time. Intel had the (superior Im sure) Pentium II and Pentium pro.

Intel brought out the Pentium 3 which started around 450MHz (Katmai core?) and eventually reached ~1.3GHz. (~700MHz to1.1GHz used Coppermine core, 1.1GHz+ used Tualatin?).

AMD brought out the Athlon to rival this. Starting at ~500MHz and in a slot package (slot A) these had 512kB of cache running at half the speed of the CPU (I think). They used the EV6 bus found in Alpha CPUs to allow the CPU to transmit data on both the rising and falling edges of the CPU FSB. This was known as DDR or Double Data Rate. These outperformed a Pentium 3 of the same clock speed. The Athlon in slot A form eventually reached 1GHz (before the Pentium 3 did) in around late 2000. Chipsets of choice were the VIA KX133 and the AMD 750 (or Irongate) chipsets. These supported PC100 SDRAM and 2x AGP. ATA66 was relatively new. Athlon Classic CPUs were made on a 0.21um process (I think. Could have been .25)

AMD then made a switch to Socket format. Pentium 3s at the time were using Socket 370 which Intel would not license to AMD. So AMD created Socket A, or more correctly, socket 462. This is one of the longest lasting sockets ever. From around mid 2000 this socket is still popular and houses CPUs ranging in speed from around 600MHz to around 2.8GHz!!

AMD created the Athlon Thunderbird as their flagship processor. This started at around 700MHz and was available in 50MHz increments to 1GHz. This used the same 100MHz DDR = 200MHz bus as the original classic Athlon did, had 128kB of Level 1 cache (vs the P3 at the time which only had 32k!) and 256kB of level 2 cache. All cache ran at the same speed as the CPU. This became known as the Thunderbird-B once the C came out. Althons kicked the Pentium 3 *** at most things, and high overclocks of ~300MHz were often achieved.

The budget CPU at the time was the Duron. This was almost identical to the Thunderbird except it only had 64kB L2 cache. Duron Spitfire CPUs were available from 600MHz to 950MHz again in 50MHz increments.

The most popular chipset at the time was probably the KT133 from VIA. This supported ATA100 (with the right southbridge) and AGP 4x. ALi and AMD also had offerings on the market but VIA was the most popular.

Once Thunderbirds reached ~1GHz AMD brought out a new revision - the Thunderbird C. This was similar to the Thunderbird-B before it but used a 133MHz DDR = 266MHz FSB. Newer chipsets were released to support it, the first of these being the VIA KT133A. This was a much better chipset than the KT133 having removed many of the bugs, and added support for 266MHz bus. The KT133/A used PC133 or PC100 SDRAM (168 pin DIMMs).

Thunderbirds and Spitfires were produced on a 0.18um process.

Now this is where Intel and AMD split. Intel designed the Pentium 4 from the ground up to scale as high as possible. And thus emerged the Willamette. This was the first of the Netburst CPUs, with 256kB L2 cache and with an L1 cache replaced with a trace cache (to cache the most recent u-ops, or microoperations, the small steps a CPU performs). It had a 100MHz FSB quad pumped to 400MHz, SSE2 (Streaming single instruction multiple data (SIMD) extentions) used the short lived Socket 423, had an enlarged pipeline to accomodate scaling as high as possible, and peformed crap. The 1.3GHz and 1.4GHz Willamettes were slower than Pentium 3 CPUs clocked at 1.0 to 1.2GHz. Willamettes ranged in speed from ~1.3GHz to ~2GHz (latter ones using a newer socket, Socket 478). Intel also designed a new Xeon CPU from the Willamette core, called Foster.These were also lacklustre performing CPUs. The chipsets at the time were the Intel i850 for desktop and i860 (Xeon) for servers. These used RDRAM. RDRAM was different from DDRRAM and ran at 800MHz. The RAM was expensive,earlier incarnations needed to be installed in pairs and any blank RIMM slots needed to be filled with dummy sticks.

AMD put the boot in to Intel with their Palomino core. Released as the AthlonXP CPU for the desktop and the Athlon 4 (being the 4th revision) for laptops this core kicked ***. Quantispeed architecture took the opposite stance to Intel - the CPUs did more each cycle and did less cycles. Support for SSE (which Intel had had since Pentium 3) was provided, the same amount of cache as the Thunderbird, and a 133Mhz DDR FSB. VIA released the KT266 chipset, one of the first to support the new at the time DDR RAM. This, like a DDR FSB,simply transferred data on both the rising and falling edges of the bus. KT266 (like any first revision VIA chipset...) wasnt too great. It allowed support for SDRAM as well and many boards appeared supporting both SDR and DDR RAM. Later,as Intel reached the end of the contract tying them to RDRAM they released the i850G (I think) which supported SDRAM rather than RDRAM. This was pathetic considering AMD had moved on to DDR RAM.

The Althon XPs were the first to use a PR rating. Instead of boasting a clock speed they had a PR rating. These started at 1500 and went up from there. An AthlonXP1500+ performed roughly as a 1500MHz Thunderbird would perform, while running at 1333MHz. This was to reflect how much more work AMD CPUs did per cycle.

The next step from AMD was to update the Duron range. AMD developed the Morgan core from the Palomino. Same FSB/cache but this scaled from 1.0GHz to 1.3GHz and added SSE support. Morgans did not overclock as well as Spitfires (some spitfires at 600Mhz were reaching >1.0GHz!!) but were decent.

Intel around this time released the Northwood core. This fixed some of the flaws of the Willamette. The process was reduced to 0.13um to alleviate heat and scaling issues and to allow the CPU to ramp to crazy speeds. To offset the longer pipeline the cache was increased to 512kB.

The Northwoods were insane overclocking chips, particularly the 1.6A and the 1.8A which were known to reach 2.5GHz+ with ease, a testement to how well these CPUs scaled.

Northwood boards tended to use SDRAM or DDRRAM, moving almost completely to the latter before the P4B emerged.

The Northwood P4s were known as the 'A' CPUs to distinguish them from Willamettes. Northwood CPUs, as well as some of the later Willamettes used an improved socket - Socket 478. Northwoods were eventually released from around 1.6Ghz to about 2.6Ghz, in the 'A' incarnation.

From the Northwood, Intel developed the Prestonia Xeon with 512kB, 1MB or 2MB cache, for server applications. They also included a new technology called hyperthreading. As the Xeons and P4s often had a lot of spare registers, and other CPU units sitting unused, hyperthreading used these spare parts to process another calculation at the same time as the main part of the CPU. This provides around a 5 to 10% performance boost in applications that suppport multithreading (many encoding, rendering and graphics packages do).

Meanwhile, AMD werent just sitting on their ***!

Next chip out the AMD fab was the Thoroughbred, or more precisely the Thoroughbred A. This was AMDs first attempt at reducing the process size to 0.13um. The chip ran OK, but didnt overclock well at all and was rather a disappointment overclocking-wise.

AMD quickly released the 'B' revision, easily one of the most famous overclocking legends. These eventually scaled from 1.47GHz right up to ~2.2GHz. These kept the same specs as the Palomino for the most part. The latter, higher speed ones, used a 333MHz FSB instead of a 266MHz FSB.

The Thoroughbred B 1700+ was an overclocking gift from the gods. These overclocked happily to 2Ghz+. In some cases 1Ghz overclocks were acheived from a 1.47Ghz CPU!! The extra layer in the process to stabilise them did the trick.

Meanwhile in the Intel camp, the Northwood was still being tweaked and updated. The Northwood B was born. This ranged from 2.4GHz (the 2.4B) to 3.06GHz, the first CPU to hit 3GHz (the 3.0B). These were almost the same as the 'A' revision but used a 133Mhz bus quad pumped to 533Mhz,rather than the Willamete/Northwood A bus at 100/400MHz. The 3.06GHz CPU was the first desktop CPU to include hyperthreading.

Going to save this and post a new post for the rest in case I slip at the last minute and lose all this!

 

[David]

So where were we? AMD Tbred Bs are kicking arse and Intel have the Northwood B doing pretty damn well!

After Tbred B AMD produced the Barton CPU. This was very similar to the Thoroughbred except this used 512kB of cache, boosting performance by an equivalent of about 150 to 200MHz. Again this used a 0.13um process.

Around week 40 of 2003, AMD did the unthinkable. Up until then, all AMD CPUs were unlocked. Thus overclockers could freely choose any multiplier they desired. This gave them an edge over Intel CPUs which had been locked since the Pentium MMX after some unscrupulous people were buying P150s and overclocking them to 166MHz and selling them on as P166 CPUs .

So from then onwards almost all new AMD CPUs were locked. The famous "pencil trick" (using a pencil lead to connect the L1 bridges on the Thunderbird and Duron CPUs to allow access to all the multipliers) didnt work. Pin modding (using small bits of wire to connect pins on the underside of the CPU) didnt work.

So overclockers looked a bit further afield and saw the XP-M or Mobile XP. Designed to run cooler than their desktop counterparts, on a 133MHz FSB rather than the 333/400MHz FSB their desktop friends used, with the same cache and optimisations, and with.... PowerNow.

PowerNow is an AMD technology for mobile CPUs where the motherboard can adjust the multiplier of the CPU to alter speed depending on how much power is needed. Translation: Hey guys! we can change the multiplier!!

Mobile Bartons on NForce 2 boards are THE AthlonXP OCing setup. These often reach 2.5GHz+ speeds.

A bit about Nforce:

Around the time of the Palominos nVidia had their first shot at a chipset - the nForce. This was an OK chipset. Then came the nForce 2 which had intergrated soundstorm (ie a **** hot integrated audio system!) the option of integrated video on par with a GeForce 4 MX (in the IGP version), built in network adapter. The NForce 2 is probably the best Socket A chipset and had loads of great overclocking options such as adjustable multpliers, FSB, CPU voltage, AGP voltage, RAM voltage. But the icing on the cake was the AGP/PCI lock.

When overclocking before, if you increase the FSB the RAM, PCI and AGP all depend on the FSB so increase as well. This led to problems when the PCI and AGP cards didnt like faster buses. Data corruption, distorted graphics and crackely sound were symptoms of PCI and AGP cards just not liking the high bus speeds. The AGP/PCI lock makes sure the PCI and AGP speeds remain constant, eliminating this issue.


The next chip out the door from Intel was the P4C. This was similar to the A and B except it boasted an 800Mhz FSB and, more importantly, Hyperthreading in all models.

After Northwood, Intel went back a bit and designed a new core, the Prescott. This was meant to allow P4s to scale further and was a horrible disappointment to start. Intel had to increase the pipeline to 30 stages from 20 to allow it to scale, and increased the cache to 1MB to compensate. Thus Prescotts are generally slower then a Northwood at the same speed. Intel also added in SSE3 (an extension to SSE2) and reduced the size of the process to 90 nanometers. This was meant to allow for higher clockspeeds but let to horrendous power leaks. CPUs were eating huge amounts of power, putting out hellish heat (hence the Presshot nickname) and were in some cases damaging motherboard power circuitry.

Prescott CPUs are denoted P4E in most cases, so the 2.8E is a Prescott CPU at 2.8GHz with an 800MHz FSB and 1MB L2 cache on a 90nm process.

The exceptions are the P4A 2.4GHz 1MB cache prescott and the P4A 2.6GHz 1MB cache prescott. These use a 400Mhz FSB (I think) and do not include hyperthreading. All P4Es have Hyperthreading (or HTT).

Other CPUs include:

- The Thornton. Failed Bartons with half their cache disabled. These are almost the same as Bartons but only have 256k cache. In some cases this extra cache can be renabled effectively giving you a cheaper Barton. Howver the extra cache may be faulty so there is no guarantee that enabling it will work.

- The Applebred. Based on the Thoroughbred B core, these are Durons which are almost the same as TbredB XPs but have just 64kB of L2 cache. These come in speeds of ~1.4GHz to 1.8GHz and have a 200 or 266MHz FSB. These tend to overclock quite well.

- The Celeron has many incarnations. There are celerons based off the Willamette core (with 128kB cache vs the 256kB on the P4s of the time) runs at ~1.3 to 1.8GHz. The Northwood based Celeron has the same specs but a speed of up to about 2.4GHz. The Celerons from around 2.6GHz on are based on Prescott and overclock exceptionally well, reaching nearly 4Ghz from 2.6 to 2.8GHz CPU!

- The P4EE or Extreme Edition is a souped up P4C with 2MB L3 cache on top of what the P4C has. This was designed to counter AMDs high end offering. These cost a LOT.

- The Nocona Xeons are based on Prescott, boast 1MB or 2MB L3 cache and have support for EM64T, Intels 64 bit extentions, in answer to the x86-64 of Athlon64/Opteron/FX CPUs.

 

[David]

Rough speed scales: plus L1/L2/bus

AMD:
Athlon Classic: 500Mhz to 1Ghz // 128/512/200
Athlon Thunderbird B: 700MHz to 1.4GHz // 128/256/200
Athlon Thunderbird C: 1GHz to 1.4GHz // 128/256/266
Athlon XP [Palomino]: 1.33GHz to 1.67GHz {XP1500+ to XP2000+} // 128/256/266
Athlon XP [Thoroughbred]: 1.47GHz to ~2.2GHz {XP1700+ to XP2800+} // 128/256/266-333
AthlonXP [Barton]: 1.83GHz to 2.2GHz {XP2500+ to XP3200+} // 128/512/333-400
AthlonXP [Thornton]: ~1.6GHz to 2.0GHz {XP2000+ to XP2800+} // 128/256/266-333
Duron [Spitfire]: 600MHz to 950MHz // 128/64/200
Duron [Morgan]: 1.0GHz to 1.3GHz // 128/64/200
Duron [Applebred]: 1.4GHz to 1.8GHz // 128/64/200-266

Intel:
Pentium 3: 500MHz to 1GHz [Katmai]
Pentium 3E: 700MHz to 1GHz [Coppermine]
Pentium 3EB: 733MHz to 1.13GHz [Coppermine]
Pentium 3-something: 1.1GHz to 1.4Ghz? [Tualatin?]
Celeron: 500MHz to 1GHz? [Coppermine]
Celeron-T: 1.0GHz to 1.4GHz? [Tualatin]
Celeron P4: 1.3GHz to 1.8GHz [Willamette based]
Celeron A?: 2.0GHz to 2.4GHz [Northwood based]
Celeron D: 2.6GHz+ [Prescott based]
Northwood A: 1.6GHz to 2.4GHz
Northwood B: 2.4GHz to 3.06GHz
Northwood C: 2.4GHz to 3.4GHz?
Prescott: 2.4GHz to 3.8GHz?

 

[c627627]

Thanks David!

Oh so there was a 1.3 GHz desktop P3 Tualatin, they didn't stop at 1.2 GHz?
I couldn't find mass released 1.3 GHz model when I looked for them:
http://www.c627627.com/AMD-Intel/

(Please let me know)

Also, there appears to have been a 650 MHz T-Bird somewhere... and those ended with the HOT 1.4 GHz Thunderbirds, remember those?


Ed Stroligo was initially incorrect. It was the Thorton, not the Thornton.
What is your stand on discussions that very few of them had bad cache. Vast majority, (almost all, it was mentioned in some discussions), were perfectly good Bartons that were disabled for marketing purposes. I'm still not clear whether piles of "faulty" cache Bartons were used or if that's an ongoing myth...

 

[c627627]

2100+ Palomino exists but only 2000+ to 2600+ Thortons, no?

 

[David]

The above three posts are not set in stone as such - they are my "best guesses".

I'm not sure about the tualatins, although I vaguely remember there being a 1.26GHz version...?

As for the 650MHz Tbirds, I never saw any. When they were released over here the slowest you could get them was 700MHz. It may be they had some 650MHz ones in the US, but over here Durons started at 600MHz and Thunderbirds at 700MHz. This was around late 2000/early 2001 when I put my first PC together, a 650MHz Duron system.

As for the 1.4GHz Tbirds, I remember them well. They were extraordinarily hot in some cases - I dont think >1.5Ghz speeds were reached without water. Temps in the 50s and 60s were the norm. Im sure Red Hot Machine (mod here) had a TbirdC 1.4 at one point, overclocked to between 1.5 and 1.6 GHz.

I had a 1.1 TbirdB myself which sat happily at mid 40s doing FaH.

As for the issue of Thorton cache problems, I thought I had read about some unsuccessful attempts at re-enabling cache. I think there may be a similar trick for Applebreds to get the exta cache back but again my recollection is too vague to say for certain.
It may be that Thortons were simply a marketing thing, although it would be a good use of Bartons with failed cache, or parts of it failing.

EDIT:
Cant remember seeing a 2100+ Pally myself, but I could be wrong.

All of the above hasnt been researched - its off the top of my head stuff.

Truth be known, Im still stuck back in the Tbird/Spitfire days. Havent had the money to keep up the hobby. I could probably overclock a Duron/Tbird system better than I could a P4E or A64

 

[c627627]

Yeah, I guess the mystery continues of just how many perfectly good Barton cores with only 1/2 cache bad were out there.

You can easily turn a Thorton into a Barton as long as it's not a late 2003 (or later) superlocked Thorton. (Superlocked Thortons cannot be modded into a Barton.)

P.S. Lots of 2100+ Palominos are owned by users at this very forum.

 

[Quailane]

EDIT:
Cant remember seeing a 2100+ Pally myself, but I could be wrong.

I owned one.