How long before Core2 becomes like Pentium4?

[Arkaign]

Don't get so detailed on me. I was speaking in general terms

Fair enough, I just think the P4 era went on for so long that a general statement cannot really be accurate. Saying it was good conflicts with the bad parts, saying it was bad conflicts with the good parts. Sigh.

 

[nenforcer]

Where are all the single core warriors hiding?

Anyone who UPGRADED to a Sempron 130/140/145 or a Celeron 430 come out of hiding and represent!

Maps props to ya and a chest bump no shame in it come out of hiding.

Core X 1 4 Life

 

[pantsaregood]

P4 sucked on launch and on closing (Prescott, Prescott2M, etc).

But if P4 Northwood sucked, then so did Athlon XP Palomino/Tbred/etc. Because the P4s were just as fast and sometimes faster. Also not that expensive unless one insisted on buying the high end models (top AXPs were also unwise chocies due to $$).

Northwood was very energy and clock inefficient. Clock for clock, it couldn't compete with the Athlon XP. It was certainly the best revision of the Pentium 4, but it wasn't exactly impressive. The original Netburst CPUs came out near the beginning of 2001, while K7 was released in mid-1999. Intel had a year and a half to optimize Netburst to outperform K7, but completely failed to do so.

 

[Arkaign]

Northwood was very energy and clock inefficient. Clock for clock, it couldn't compete with the Athlon XP. It was certainly the best revision of the Pentium 4, but it wasn't exactly impressive. The original Netburst CPUs came out near the beginning of 2001, while K7 was released in mid-1999. Intel had a year and a half to optimize Netburst to outperform K7, but completely failed to do so.

Hmm. Clock for clock wasn't really the game with the P4, but $ for $ it made sense to buy lower clocked Northwoods or AXPs and overclock them.

And in fact, the Northwood left behind the AXP several times, the last time for good. Anand's own comparison of the 3200+ showed it to be completely outgunned by the 3.0c even.

 

[pantsaregood]

Hmm. Clock for clock wasn't really the game with the P4, but $ for $ it made sense to buy lower clocked Northwoods or AXPs and overclock them.

And in fact, the Northwood left behind the AXP several times, the last time for good. Anand's own comparison of the 3200+ showed it to be completely outgunned by the 3.0c even.

By the time the Pentium 4 HTs were released, K8's release was impending. There were only a few months between the release of Pentium 4 HT and Athlon 64.

Athlon XP and Northwood B were pretty even generally. By the time scaling became an issue for K7, K8 was ready.

 

[Arkaign]

Hyper-Threading P4s became available in November of 02, though indeed the 800mhz FSB 'C' revision didn't release until April of '03.

The K8's initial release in September of '03 was problematic, for several reasons (at least on the desktop front, for workstations it was much less so).

It launched first with Socket 940, which demanded special ram, and the mainboards were quite expensive. The clawhammer 3200+ on socket 754 was a pretty even competitor for the P4 3.2, beating it in more than it lost imho, but we all know how socket 754 turned out.

In a weird way, Intel became the value leader during this time, as you could get a cheap Socket 478 setup, use dual-channel regular DDR, and overclock a cheap Northwood to ~3.2ghz or a little more, wheras the AMD64 chips and especially socket 940 mobos/ram, were quite expensive and uncommon.

Where AMD finally started getting good traction was in June of 2004, when they launched socket 939, eliminating the need for buffered DDR, and simultaneously launched Winchester. The doubling of memory performance, increase in HT from 800 to 1000, and overall affordability (it now made perfect sense for new buyers to grab a s939 board and a midrange A64 chip) made it begin it's march to total leadership. P4s were still competitive in performance, but this is where I think it was hands-down the better choice to go AMD64.

Things are a lot more complex than people commonly remember. K8 might have been released only 5 months after the P4C, and 10 months after the first hyper-threading P4s, but it was really about a year to a year and a half before AMD's really started going places with AMD64, eventually leaving P4 in the dust completely.

 

[Munky]

C2D will beat those quads for everyday performance.
http://www.anandtech.com/bench/Product/87?vs=106


An E7500 is 50% faster than my X2 5200+, yet I doubt I'll see the need to upgrade in the next 3-5 years.

Like what, web browsing? For any actual work that requires a fast CPU, like 3D rendering, software compilation, or movie encoding, the Athlon2 X4 quads will beat whatever your C2D can muster. See http://www.behardware.com/articles/768-10/amd-athlon-ii-x4-620-and-630.html

 

[exar333]

Hyper-Threading P4s became available in November of 02, though indeed the 800mhz FSB 'C' revision didn't release until April of '03.

The K8's initial release in September of '03 was problematic, for several reasons (at least on the desktop front, for workstations it was much less so).

It launched first with Socket 940, which demanded special ram, and the mainboards were quite expensive. The clawhammer 3200+ on socket 754 was a pretty even competitor for the P4 3.2, beating it in more than it lost imho, but we all know how socket 754 turned out.

In a weird way, Intel became the value leader during this time, as you could get a cheap Socket 478 setup, use dual-channel regular DDR, and overclock a cheap Northwood to ~3.2ghz or a little more, wheras the AMD64 chips and especially socket 940 mobos/ram, were quite expensive and uncommon.

Where AMD finally started getting good traction was in June of 2004, when they launched socket 939, eliminating the need for buffered DDR, and simultaneously launched Winchester. The doubling of memory performance, increase in HT from 800 to 1000, and overall affordability (it now made perfect sense for new buyers to grab a s939 board and a midrange A64 chip) made it begin it's march to total leadership. P4s were still competitive in performance, but this is where I think it was hands-down the better choice to go AMD64.

Things are a lot more complex than people commonly remember. K8 might have been released only 5 months after the P4C, and 10 months after the first hyper-threading P4s, but it was really about a year to a year and a half before AMD's really started going places with AMD64, eventually leaving P4 in the dust completely.

Thats a great summary.

My own personal; upgrade path followed this quite closely. I was in school and money was tight, so bang-for-the-buck was key for me. I rocked the 2800+ (OCd at ~2.1ghz) for a pretty long time until 939 came out. I grabbed a 3200+ Winnie and OCd that to about 2.5ghz and that lasted me for a long time until a cheap C2D pentium e2160 build. I lasted on that until Nehalem in '08, which I have been happily running since.

I probably do not need a new computer, but after getting so used to building them regularly from 1995-2008, I am really getting the upgrade bug now. I am sure when I pass this on to someone in my family (niece or nephew maybe?) they will enjoy the computer for some time.

 

[ElFenix]

probably not until the next round of consoles comes out. for most games it's still satisfactory. for video encoding quads aren't really beaten by the i3s (and video encoding takes long enough that it's something done while AFK anyway). for using the internet i doubt anyone could tell any difference.

if your upgrade choice is between getting an i3 (w/ram and mobo) or getting an SSD for a c2q system, the SSD is the better choice.

 

[BonzaiDuck]

All this discussion of "number of cores" versus desired performance versus adequate performance has to address market segments. I see a pile of people still hamstrung by their comfort-zone affinity to Win XP-32 -- with hardware that is like a racehorse still held back at the starting-gate. Then there are the mainstreamers who want performance for a low price, and the generational improvements by Intel or AMD from one year to the next result in various manufacturing efficiencies and product differentiation within that technology to make the "great price" feasible.

Then . . . there are the enthusiasts and over-clockers.

I'm dumping my Q6600/680i system because the mobo has limitations. I might have lived with those limitations for another three years or more -- and still had better than "mainstream" performance if you don't consider the enhancements of the latest CPUs and chipsets. But it then becomes a choice between LGA775 architecture versus Nehalem/Clarksdale/SAndy-Bridge.

For this, I'll pay out a net amount somewhere in the $1,500-$1,700 range -- excluding the computer case. This occurs because I choose a $200 mobo over $100; OEMs get their Windows-7 bundle for pennies while resellers price the surplus discs/licenses $30 below retail-box but $100 more than OEM customers will pay for a bundled OS; I'll spend at least $400 for storage technology while the OEM solution would cost less than $100 just for retail; and then there's the "graphics factor."

Compare this to a "user" who spends $700 on a Dell built with an AMD hex-core Phenom-X6, and another $300+ on warranty and extended service contract.

OEM sales people report that mainstreamers are ordering their machines without PCI-E graphics cards -- opting for on-board graphics solutions. They're making decisions where price overwhelms performance desires, and the onboard graphics are adequate for a statistical majority of buyers.

My friend bought his XPS 7100 because his 5-year-old Dell went south with mobo and memory damage. Otherwise, he might have waited another year, two -- maybe more. Who knows?

I already said somewhere that my E8600@4.1+Ghz seemed "fast," but now that I've passed the milestone of firing up my Z68 build for the first time, made a dry-run installation of the OS while I wait for a VelociRaptor to arrive Friday -- I haven't even OC'd the i7-2600K and there's just . . . no comparison.

The old Wolfdale system could go another three years, still handle all the games I could throw at it, and meet all needs. I should sell it, but can't bear to out of some sort of fondness; I could just give it away to a family member, but it's beyond their needs or appreciation. It's my toy . . . .

Even mainstreamers, though, will get snagged by the "wants" hook. Another friend had a four-year-old AMD dual-core system infected with malware, and she won't spend time -- nor does she have the experience -- to clean it up on her own barring a complete reinstall of the OS and all software. I actually thought she might just go out and buy a new machine for panic hinged to impulse -- which I'd say would be a mistake unless it was just based on "want" for the very latest technology. She finally opted to pay some geek squad outfit to fix things for $140. She won't play any games; doesn't even use Media Center. That machine should be good for a while . . . .

 

[pantsaregood]

Hyper-Threading P4s became available in November of 02, though indeed the 800mhz FSB 'C' revision didn't release until April of '03.

The K8's initial release in September of '03 was problematic, for several reasons (at least on the desktop front, for workstations it was much less so).

It launched first with Socket 940, which demanded special ram, and the mainboards were quite expensive. The clawhammer 3200+ on socket 754 was a pretty even competitor for the P4 3.2, beating it in more than it lost imho, but we all know how socket 754 turned out.

In a weird way, Intel became the value leader during this time, as you could get a cheap Socket 478 setup, use dual-channel regular DDR, and overclock a cheap Northwood to ~3.2ghz or a little more, wheras the AMD64 chips and especially socket 940 mobos/ram, were quite expensive and uncommon.

Where AMD finally started getting good traction was in June of 2004, when they launched socket 939, eliminating the need for buffered DDR, and simultaneously launched Winchester. The doubling of memory performance, increase in HT from 800 to 1000, and overall affordability (it now made perfect sense for new buyers to grab a s939 board and a midrange A64 chip) made it begin it's march to total leadership. P4s were still competitive in performance, but this is where I think it was hands-down the better choice to go AMD64.

Things are a lot more complex than people commonly remember. K8 might have been released only 5 months after the P4C, and 10 months after the first hyper-threading P4s, but it was really about a year to a year and a half before AMD's really started going places with AMD64, eventually leaving P4 in the dust completely.

"Pentium 4 HT" is a branding. The 3.06 GHz Pentium 4 didn't fall under that branding.

In what ways was K8 problematic to begin with, and why the negative take on socket 754? It may have been a short-lived socket, but AMD kept it at least somewhat updated by releasing Venice models for it. Chipsets on 754 generally stayed up to date with their 939 counterparts. I recall 754 being heavily recommended as a gaming platform in its time.

 

[B-Riz]

In essence, the Conroe architecture is a super duper P3, developed from P3 mobile because P4 was not good for mobile applications.

The Pentium M was Intel's best CPU until C2D was released, and only available in mobile computers.

http://en.wikipedia.org/wiki/Pentium_M_(microarchitecture)#Pentium_M

Everyone talking about # of cores should seriously think about how far behind software is compared to the hardware available.

Thinking and programming in multiple threads is not exactly easy compared to the x number of years single threaded programming / development has been used.

The O/S will have a bigger impact on the "feel" of using the computer than a tick/tock cpu upgrade.

And please take into account L2 and L3 cache sizes when comparing cpu's.

C2D is setup so the more cache it has, the better it performs.

The bigger question is why does Intel have 4 consumer sockets now?

I would say corporate computer life cycle is the only reason 775 is still around, and will stick around. It has been proven stable, and more than fast enough for the MS suite of software.

But, 1155 is a way better deal now for the consumer, and with ram and hdd prices at the lowest ever, it is nice to get an awesome computer for a few bills.

I'm still pissed I wasted the money on a i7 920 1366 platform, which was just something to dump a server cpu on the enthusiast, when it offered minimal improvement over the C2Q in relation to the software and games at the time.

Nothing after Core 2 mattered until Win 7 came out and could use gobs of ram.

 

[piasabird]

The failure of the P4 due to overheating, lead to an AMD Era. What lead to this was a laze' faire attituted of the P4 being fast eneough and that the consumer did not need 64 bit processing. Since then, it is like Intel has learned a lesson because faster and faster sells. However, they also learned that sometimes small and mininimalistic sells with the Atom processor. So they started working it from both ends.

So Intel has already gone to 2nd generation Core 2 Duo; however, they still sell pentium processors.

So how long till we get a 128 bit processor?

One way to improve computing is to drastically improve storage technology or increase the write spreed on drives or use multiple platters and spread the data out further. Maybe a better hybrid drives system where they only use the outer half of the disk to write data would help a bit.

 

[pantsaregood]

The failure of the P4 due to overheating, lead to an AMD Era. What lead to this was a laze' faire attituted of the P4 being fast eneough and that the consumer did not need 64 bit processing. Since then, it is like Intel has learned a lesson because faster and faster sells. However, they also learned that sometimes small and mininimalistic sells with the Atom processor. So they started working it from both ends.

So Intel has already gone to 2nd generation Core 2 Duo; however, they still sell pentium processors.

So how long till we get a 128 bit processor?

One way to improve computing is to drastically improve storage technology or increase the write spreed on drives or use multiple platters and spread the data out further. Maybe a better hybrid drives system where they only use the outer half of the disk to write data would help a bit.

2nd generation Core 2 Duo? All Core 2s (Solos, Duos, and Quads) were Conroe-based. There was no second generation.

I also don't know that we'll ever see 128-bit computing. The amount of memory a 128-bit bus could address is absurd. At 2^128 bits of addressable RAM, you're running into a point at which we lack the raw materials to meet such capacity.

 

[sangyup81]

Ask the developers of Adobe Flash. P4's and single core Athlon 64s would still be somewhat usable if it weren't for them

 

[pantsaregood]

Ask the developers of Adobe Flash. P4's and single core Athlon 64s would still be somewhat usable if it weren't for them

"Somewhat usable" is an understatement. Adobe's products can make a new PC run like trash. Adobe Reader doesn't scroll smoothly on my 2500K and 8GB of RAM. Flash just bogs down old PCs.

 

[dave_the_nerd]

Ask the developers of Adobe Flash. P4's and single core Athlon 64s would still be somewhat usable if it weren't for them

Preach it, brother!

 

[Arkaign]

"Pentium 4 HT" is a branding. The 3.06 GHz Pentium 4 didn't fall under that branding.

In what ways was K8 problematic to begin with, and why the negative take on socket 754? It may have been a short-lived socket, but AMD kept it at least somewhat updated by releasing Venice models for it. Chipsets on 754 generally stayed up to date with their 939 counterparts. I recall 754 being heavily recommended as a gaming platform in its time.

I'm aware of the branding, but that's kind of irrelevant to the issues here.

K8 was problematic to begin with much in the way that P4 was problematic to begin with, though imho a bit less so, in that it launched with a socket that required expensive mobos and expensive memory that didn't carry over from people's existing builds. Socket 940 required buffered DDR

Socket 754 was irritating because it was so lazy. They could have more easily prepped Socket 939 to be ready to go from the outset, as there were only the most minor differences between 939 and 940 in terms of how it all worked, it was only the support for multi-socket and buffered or registered DDR that differentiated it. Limiting it to single-channel memory, and keeping it from supporting the forthcoming dual-core chips was a poor decision imho. They already had a good 'value' line with the AXPs that were still produced, it seemed to overlap for no good reason. I would have still recommended a S754 build over say a P4 3.2C, but I'd have recommended a P4 2.4 or AXP 2500+ over either of them simply due to the cost effectiveness. AMD could have offered a A64 2500+ or so for a value option, but they didn't, though they did eventually release some crippled 128k cache Semprons. With either the cheaper P4 or AXP you could get A64-3000+ S754 performance after overclocking, and have much more $$ left over for GPU, memory, etc. Once socket 939 hit, however, and the excellent NF4 boards, it was no contest, S939 FTW.

Anyway, it was a stark contrast to the simplicity that AMD kept in the K7 era. Socket A was stable for a very long time, and even some early KT133a mobos could be flashed to run relatively powerful AXPs as they came out over time. Nforce2 brought a lot to the table as well.

 

[WhoBeDaPlaya]

2nd generation Core 2 Duo? All Core 2s (Solos, Duos, and Quads) were Conroe-based. There was no second generation.

1st gen = Conroe = 65nm
2nd gen = Wolfdale = 45nm

 

[Idontcare]

2nd generation Core 2 Duo? All Core 2s (Solos, Duos, and Quads) were Conroe-based. There was no second generation.

1st gen = Conroe = 65nm
2nd gen = Wolfdale = 45nm

I understand that we forum-rats generically refer to 45nm wolfdale (the tick to Conroe's tock) as the "2nd gen" but pantsaregood is technically correct in asserting that there is/was no such thing as a "2nd generation Core 2 Duo" per Intel's definition of its product families:

http://www.intel.com/products/processor_number/about/core.htm

(Core 2 Duo was not simply based on Conroe though, as WhoBeDaPlaya rightly points out, but it is not correct to refer to Wolfdale as "2nd gen" either)

Chronologically it goes:

1. Intel® Core 2 Duo (Conroe/Wolfdale) & Intel® Core 2 Quad (Kentsfield/Penryn)
2. Intel® Core (Nehalem/Westmere)
3. 2nd generation Intel® Core (Sandy Bridge, and presumably Ivy Bridge)

They combine ticks and tocks in their brand naming.

It took me a while to figure out what the heck made Sandy Bridge worthy of being labeled "2nd gen" whereas Nehalem was not "2nd gen".

 

[WhoBeDaPlaya]

Aaah, glad we cleared up the nomenclature. At least it didn't result in a Russian fellow frantically yelling "Nyet nyet Houston! It's metric!"

 

[B-Riz]

Penryn technically was a second generation Core 2 Duo.

It had micro-architecture improvements along with a die shrink.

http://download.intel.com/technology/itj/2008/v12i3/paper2.pdf

Wolfdale was just a codename.

Merom -> Penryn

 

[Ken g6]

Chronologically it goes:

1. Intel® Core 2 Duo (Conroe/Wolfdale) & Intel® Core 2 Quad (Kentsfield/Penryn)
2. Intel® Core (Nehalem/Westmere)
3. 2nd generation Intel® Core (Sandy Bridge, and presumably Ivy Bridge)

What's confusing about that is that, before any of those processors, there was the Intel® Core™ Duo.

 

[toyota]

What's confusing about that is that, before any of those processors, there was the Intel® Core™ Duo.

lol, and the Core Duo was basically just as fast clock for clock as the Core 2 Duo. whats sad is my friend has a sub $600 notebook from 5 years ago with a 2.0 Core duo and he would have to spend almost that much today just to beat it. same goes for my parents $300 comp from 3 years ago. they would have to spend more than that now just to have a faster system than that. I sure dont see all this progress in lower end stuff people talk about.

 

[VirtualLarry]

Does anyone think that Bulldozer's 8 cores will put the nail in the coffin of the dual-core Core2Duos, totally obliterating it in multi-tasking power?