- Aug 10, 2001
- 10,420
- 2
- 0
The design relied heavily on clock frequency, yet Intel couldn't get the Pentium 4 anywhere near the frequencies they promised. Didn't they once claim it could be scaled up to 10+ GHz?
Was the Pentium 4 the worst processor design in history?
- Thread starter Random Variable
- Start date
WRONG FORUM!!!
Edit: Maybe not the worst but it was pretty bad. Considering that the Pentium 3 was actually better.
Edit: Maybe not the worst but it was pretty bad. Considering that the Pentium 3 was actually better.
- Aug 10, 2001
- 10,420
- 2
- 0
dullard
Elite Member
- May 21, 2001
- 25,761
- 4,282
- 126
There were a couple of times when it had the best performance of any home-user processor ever made (by any company). So, how could that possibly be considered the worst design?
Sure, it was more often than not a bit behind AMD, but it was certainly in the running most of the time. And after 1.6 GHz on up it was faster than all P3s in virtually all tasks.
Sure, it was more often than not a bit behind AMD, but it was certainly in the running most of the time. And after 1.6 GHz on up it was faster than all P3s in virtually all tasks.
It was only faster than the PIII past 1.6ghz because they didn't make any PIIIs past 1.4ghz... Had they done so, the P4 would remain inherently inefficient in relative terms. Looking at the athlon Xp and the Intel Pentium M makes you realize that the architechtures were highly inefficient, including the Pentium III. They're able to achieve significantly more per clock cycle than ever before and makes you realize how much you were missing out on.
Its quite sad really, though I do have to wonder, if they knew now what they did back in '99/'00 with the exception of achieving smaller micron processes, would they have been able to achieve the performance of today's Pentium M processor? Maybe even core duo? (Clock for clock of course...) Or, how big of a performance gap or how close would they have gotten to the clock for clock performance of the Pentium M or Core duo clock for clock?
Actually, if I had to choose between a P3 1ghz/1.4ghz tulatin or a P4 1.5Ghz/1.6ghz machine, I'd choose the PIII, the P4s of that era were sluggish, I know because I have a few of them.
My old 1.6a did 2.71ghz
My old 2.4b did 3.33ghz
My old 2.4c did 3.5ghz and had Hyperthreading which gave up to 20% gains in apps that were multithreaded
They got better and faster once the socket changed to 478 and they became northwoods....Once they hit the 90nm debacle and the prescotts it started going downhill until it thudded with the crap known as smithfield.
My old 2.4b did 3.33ghz
My old 2.4c did 3.5ghz and had Hyperthreading which gave up to 20% gains in apps that were multithreaded
They got better and faster once the socket changed to 478 and they became northwoods....Once they hit the 90nm debacle and the prescotts it started going downhill until it thudded with the crap known as smithfield.
I wasn't into tech stuff much back when the Pentium 4 first launched, but I do remember being confused by people saying to me that the Pentium 4 was faster but clock for clock it's slower.
It wasn't necessarily a bad design, its just that it first relied too much on optimisations (like SSE2, and later Hyperthreading) and high click speeds. Even though Pentium 4 was beating AMD during the Athlon XP vs Northwood days, I think that was down to AMD having delays in getting A64 launched.
Overall in terms of pure performance, it still rivalled the competition and it has made Intel more aware of potential problems when designing future CPUs.
It wasn't necessarily a bad design, its just that it first relied too much on optimisations (like SSE2, and later Hyperthreading) and high click speeds. Even though Pentium 4 was beating AMD during the Athlon XP vs Northwood days, I think that was down to AMD having delays in getting A64 launched.
Overall in terms of pure performance, it still rivalled the competition and it has made Intel more aware of potential problems when designing future CPUs.
they should really have got rid of netburst after they released northwood or atleast the intial prescotts.But they thought the die-shrinks would solve the heat and speed issues but didn't.
But i guess alls well that ends well.
Now they have the core architecutre and they are making up for the sins of the past.And are not gonna hang onto a architecture for more than 2 yrs.All good for the consumer.
Netburst was inefficient but not the worst thing ever.Infact most companies and homes around the world is full of pentium 4/D.
But i guess alls well that ends well.
Now they have the core architecutre and they are making up for the sins of the past.And are not gonna hang onto a architecture for more than 2 yrs.All good for the consumer.
Netburst was inefficient but not the worst thing ever.Infact most companies and homes around the world is full of pentium 4/D.
Heck no! It's a lot better of a design than the: 8088, 8086, 286, 386, 486, PI, and PII, for starters. Later models (Northwood and beyond) were better than the PIIIs, also.
Apparently it depends on what your definition of "better" is - faster? Sure. Architecturally? Maybe not. Raw power/clock cycle? Definitely not.
Apparently it depends on what your definition of "better" is - faster? Sure. Architecturally? Maybe not. Raw power/clock cycle? Definitely not.
To answer the question, not even close...Netburst was an excellent design.
Intel did make the prediction that they would be at 10GHz by now, but they never said it was to be the P4.
To be fair to Intel, it just wasn't possible to predict that leakage would occur at the rate it ended up doing so. If the leakage curve had been closer to linear (instead of being almost exponential), then they probably could have done it...
That said, the real issue isn't how good or bad Netburst was, it's the fact that it was necessarily compared to AMD's A64. The K8 (and K7 to a great degree) were unpredictably brilliant chips...I consider the innovation of their design to be on par with the Intel PPro (another milestone in chip architecture).
I would personally call it the most successful failed processor design. It was inevitable that someone would try a netburst like architecture eventually given the competition between AMD and Intel. However, Intel failed to foresee the heat issues and leakage caused at those high clocks.
If netburst had been as successful as they originally thought it would be AMD would probably not exist (10 GHz netburst chip > anything AMD has). Intel made a mistake, but they also showed conclusively for the first time that clock speeds cannot be increased forever. In that respect I would say it was successful.
Now you want to talk about failure look at the Itanium that was a failure.
If netburst had been as successful as they originally thought it would be AMD would probably not exist (10 GHz netburst chip > anything AMD has). Intel made a mistake, but they also showed conclusively for the first time that clock speeds cannot be increased forever. In that respect I would say it was successful.
Now you want to talk about failure look at the Itanium that was a failure.
Weeeellll, actually all they showed was that it can't be done with current technology.
To give you an example of what I mean, carbon nanotubes have up to 1,000-10,000 times the conductivity of copper...
So it's quite possible that we COULD see 50GHz chips in the next decade...
As to Itanium, that was purely a marketing failure...Intel made a very bad call as to the willingness of companies for converting to all new software, and developers for spending Billions developing the software on spec. The engineering of Itanium is brilliant, it was the assumption by Intel's management that was flawed (greatly!). They also didn't count on Opteron and AMD64 either...
Sigh, the good 'ol days...
http://www.chip-architect.com/news/2002_04_16_Prescott_Prospects.html
lol, SMT bringing 50-60% extra performance, 6ghz on a 100nm process... :laugh:
http://www.chip-architect.com/news/2002_04_16_Prescott_Prospects.html
lol, SMT bringing 50-60% extra performance, 6ghz on a 100nm process... :laugh:
The 3.8 GHz P4 and 3800+ A64 were very close in performance if you exclude gaming. Really, if you were doing encoding, the P4 would have probably been the faster choice. Of course, the P4 took more power and produced more heat. The A64 was a great chip but it really didn't revolutionize performance over the P4. The FX chips of the time were faster than any P4 but it cost way too much to be mainstream. It wasn't until the release of the X2 that AMD took the clear performance lead in the performance AND enthusiast market segments.
Of course, that was forfeited with the release of the Core 2 which really does have revolutionary performance in the mainstream, enthusiast, and performance segments. It's pretty clear that Intel could release a 3.2-3.4 GHz Conroe right now and destroy anything AMD has in the near to mid-term future. I hope that AMD implemented real architectural improvements in the K8L because it won't be long until Intel sticks an IMC on the Conroe and really lets it fly. I'm also nervous because Intel has really been delivering as of late whereas AMD has been falling behind. I would say Intel sat too long on the P4 core and was clearly beaten by the X2s but it looks as though AMD has sat far too long on the K7 and is clearly beaten by Core. No matter how many cores they can pack on a die, it won't eliminate the fact that the Core microarchitecture is just better.
Of course, that was forfeited with the release of the Core 2 which really does have revolutionary performance in the mainstream, enthusiast, and performance segments. It's pretty clear that Intel could release a 3.2-3.4 GHz Conroe right now and destroy anything AMD has in the near to mid-term future. I hope that AMD implemented real architectural improvements in the K8L because it won't be long until Intel sticks an IMC on the Conroe and really lets it fly. I'm also nervous because Intel has really been delivering as of late whereas AMD has been falling behind. I would say Intel sat too long on the P4 core and was clearly beaten by the X2s but it looks as though AMD has sat far too long on the K7 and is clearly beaten by Core. No matter how many cores they can pack on a die, it won't eliminate the fact that the Core microarchitecture is just better.
I dunno, dude, the Pentium D 805 is pretty freaking bad... at least my friends is. He has to cool the crap out of it, and he doesn't OC at all. It's also quite slow, as is my 820. Mine doesn't run nearly as hot as his though.
While I do partially agree with you, because of intel's cheap implimentation of dual core via the FSB onto their processors, while they may be fine with the current speeds, as they add more cores, there will be a significant drop off in performance improvements as the FSB will become a severely limiting factor. For the AMD platform and the HTT, the links between the two processors and the board its self won't be an issue and should scale significantly better.
Not a chance. Merced (first gen Itanium) and US3i (SPARC) are examples of fairly modern processors that are FAR worse in terms of performance.
It wasn't better, it couldn't clock anywhere near as high. Even the Pentium M, a heavily modified Pentium 3 (and thus a different architecture) couldn't perform as well as the P4 in many tasks, primarily the ones the P4 was pushed for like 3d rendering and photo editing.
I'd say the P4 was a pretty good design that ran into some unexpected engineering difficulties. I'd consider the Athlon a better design though, with far worse manufacturing on AMD's side.
Only because P4 ran into some unexpected initial problems (it was a very ambitious design) and was released at lower clock speeds than expected, and with slower ram the expected. (sdr instead of rambus)
Anyhow, the P3, like the G4, did not clock very high and didn't have high bang per mhz. P-M improved in many areas but was still lacking in clock speed and bang per mhz in some tasks.
And the inefficiency of a P4 really only matters in die size and power consumption, which generally weren't that bad until Prescott.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 24K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 19K
-
-
-
Facebook
Twitter