PDA

View Full Version : Prescott to have 30 stage pipeline?


Pages : [1] 2 3 4

Viral
01-22-04, 08:25 AM
http://news.com.com/2100-1006_3-5144907.html?tag=nefd_top

This is bad news if true IMO.. it means intel are up to their old tricks again and plan on trying to fool everyone into thinking MHz is all that matters.

It is not confirmed that the pipeline will be made of of 30 stages, compared to the current P4's 20 (athlon 64's 12), but it is confirmed that "It will have a larger pipeline".

This would be good news if intel planned of scaling the processor much higher in clock speed.. but for me to be happy with MHz compensation for such a pipeline length increase, they would have to launch prescott at 3.8GHz, something they wont be doing.

Sure alot more people now are weary about what determines a processors speed.. but thats just the thing.. those other cases where more is better will also improve on the prescott (more cache, SSE3, improved HT). However, what many will not know when buying such a 30stage pipeline processor, is that it will likely be slower clock for clock than a Northwood. That is at least until SSE3 apps rock up, and even then, who knows.

For those who don't know, intel did this in the past increasing the P4's pipeline to 20stages from the P3's 10. This meant a P3 at 1GHz would often outperform a 1.4GHz P4, and a 1.2GHz Athlon would give it a thrashing (even with it's then 15 stage pipeline).

A longer pipeline also means more dependancy on cache due to errors taking longer to be rectified due to the longer pipeline. This means even with 256KB L2 cache, a prescott celeron will most likely perform as dismally as todays celerons. They will have twice the cache, but a 50% longer pipeline.

So really.. who are intel fooling? joesixpac most likely:D well i'm sure this will allow them to get more of joe's money. Will it allow them to get yours?

Nutty
01-22-04, 09:03 AM
I heard rumours that Intel will have to adopt a PR rating system. Something they've slagged off in the past, just to convince the public the chips are that much faster.

IMO, longer pipelines with relying on high clock speeds for performance is a bad idea. You can only go soo fast..

I much prefer AMD's more power per clock philosophy.

Ninja Prime
01-22-04, 12:03 PM
More complete retardedness from Intel, reminding why I haven't bought an Intel processor in 7 years. Can you even believe this quote from one of their own reps:

"It will have a larger pipeline," the representative said. "The larger the pipeline, you can do less work per clock and speed up the processor."

....Soooo he's saying they are going to give us a slower/equal processor that says "4ghz" on the side, and sell it to us for a top of the line price? MorAns. Its the same BS from the original P4 all over again...

Bob: "Hey Jim, those AMD guys are kicking our ass again, what do with do?"

Jim: "Lets make the pipeline longer again, then we can clock it 25% faster, but we'll take a 30% per clock performance hit, and we can show all the morons out there a pretty 4ghz and they'll think its fast!"

Bob: "That's why you're the boss!"

Sazar
01-22-04, 01:27 PM
I'll ask my brother if he has worked with any of these and give some sort of response tonite if I remmeber...

Dazz
01-22-04, 01:36 PM
It makes sence, the P4 was slow when it first came out now look at it! It means they can boost the peformance further with just extra cache and bandwith, AMD CAN NOT, dual channel doesn't even work for the Athlon 64 FX in real world applications. You poeple are just nit picking.

Sazar
01-22-04, 01:41 PM
Originally posted by Dazz
It makes sence, the P4 was slow when it first came out now look at it! It means they can boost the peformance further with just extra cache and bandwith, AMD CAN NOT, dual channel doesn't even work for the Athlon 64 FX in real world applications. You poeple are just nit picking.

amd can't boost performance with extra cache :confused:

um... 3000+ and 3200+ difference is cache init ?

newcastle has 1/2 the cache...

Dazz
01-22-04, 02:04 PM
Yeah but the 3200+ can't even compete with Intels P4 3.2C Also the diffrence between the 3000+ & 3200+ was 200MHz or 66MHz + 512K cache more.

Sazar
01-22-04, 02:08 PM
Originally posted by Dazz
Yeah but the 3200+ can't even compete with Intels P4 3.2C Also the diffrence between the 3000+ & 3200+ was 200MHz or 66MHz + 512K cache more.

3000+ and 3200+ both retail @ same clock speed @ newegg with difference just being cache...

Nutty
01-22-04, 03:39 PM
I'm not convinced its the best route to take. At less work per clock, you have to provide bigger clock speed increases, just for the similar performance increase.

Say Processor X runs half the speed of Y, but has similar performance level.

If Processor X has a 100Mhz increase, Processor Y will require 200Mhz increase to stay the same.

Theres also the fact that at really high speeds, system memory becomes an increasing bottleneck in the system, which is why Intel are having to double the cache again.

It just doesn't add up to a very efficient solution IMO.

Malfunction
01-22-04, 03:55 PM
I have to agree, it does not make sense. Guess this year I will return to AMD processors and hope they improve apon their exisiting line. I hope the Multi-tasking benefits that Intel has with Hyper-Threading exist in some form on the AMD side. I would surely miss that feature alone if there is nothing similar being offered. :(

Peace,

:afro:

Viral
01-22-04, 09:49 PM
Originally posted by Dazz
Yeah but the 3200+ can't even compete with Intels P4 3.2C Also the diffrence between the 3000+ & 3200+ was 200MHz or 66MHz + 512K cache more.

No he is talking about the A64 line.. both are clocked at 2.0GHz.

What you are saying doesn't make sense.. what is the benefit of having an architecture that benefits from more cache and a faster FSB if these higher benefits are just making up for the lost performance from elongating the pipeline in the first place?

Also.. implementing more cache will up the production cost (and besides intel arn't planning to increase their L2 cache for some time after prescott), and upping the FSB means more expensive memory solutions are required (among other things), a dependancy on upping these two things is not a cost effective solution to gaining performance. An efficient processor that does not rely heavily on external(and other internal, ie cache) elements is the best cost efficient way to increase performance IMO.

Basically intel are planning to ship prescotts with lower IPC (instructions per clock) and hope that the cache, HT and SSE3(eventually) will make up for it. Unfortunately, with a 30 stage pipeline, i dont see this happening.. and it is likely a 3.4GHz prescott will not outperform a 3.2GHz P4 at all in many(most?) apps. So are they going forward at all? Eventually yes. However to most consumers, they will be going forward like usual (GHz).

Afew months ago i was expecting prescott to have a higher IPC than Northwood.. just goes to show intel arn't changing from their old ways any time soon. I am very dissapointed in this, and hope they get what they deserve for thinking they can fool everyone with this. Seriously, I'm not trying to be a fanboy, but who can't see their real intentions behind this?

Malfunction
01-22-04, 10:07 PM
Well, if they are stealing the lower end segment with price dropping and Multi-Threading performance, then I will say they are going to win by a huge margin.

The great thing about this is no one really has a clue about the strategy. So for the most part, it is purely speculation. I mean seeing these prices of Intel CPU's drop like crazy makes me very suspicious of their tactics.

I mean... come on, a 2.8C for $163?!?! That is crazy cool!

Peace,

:afro:

druga runda
01-23-04, 12:31 PM
I'm afraid that AMD64 with 64bit windows @2.6 ghz (by year end)... will have to have 5000+ PR rating now (and will still be 20% faster with most games/business apps)

I know Intel is trying hard to discredit AMD's PR rating sheme with all this... :p

bkswaney
01-23-04, 04:30 PM
For us overclockers this is great.
Longer pipes = higher clocks. :D
Even if it comes out at 3.4 if it can hit 4+ghz I'll be most happy. ;)

StealthHawk
01-23-04, 04:49 PM
Originally posted by Ninja Prime
More complete retardedness from Intel, reminding why I haven't bought an Intel processor in 7 years. Can you even believe this quote from one of their own reps:

"It will have a larger pipeline," the representative said. "The larger the pipeline, you can do less work per clock and speed up the processor."

At least they're being honest.

Nutty
01-23-04, 05:06 PM
For us overclockers this is great.
Longer pipes = higher clocks.
Even if it comes out at 3.4 if it can hit 4+ghz I'll be most happy.

But at 4Ghz, it'll prolly only match a 3.2P4, or 3400+ AMD in performance, and tbh I doubt we'll see past 4Ghz this year. Ppl always went on about P4's strength is in its speed, and how we'll have 5Ghz P4's. It didn't happen, it isn't gonna happen, and I doubt the new chip will get very high anytime soon either.

Viral
01-24-04, 12:42 AM
Originally posted by Nutty
But at 4Ghz, it'll prolly only match a 3.2P4, or 3400+ AMD in performance, and tbh I doubt we'll see past 4Ghz this year. Ppl always went on about P4's strength is in its speed, and how we'll have 5Ghz P4's. It didn't happen, it isn't gonna happen, and I doubt the new chip will get very high anytime soon either.

Exactly, it's all relative.. You get a higher overclock, so you think you are getting better performance, but really you are getting nothing extra at all.

It's the same as overclocking a 2.0GHz A64 to 2.2GHz and a 3.0GHz P4 to 3.3GHz.. the percentage increase is exactly the same. The performance increase.. well i would say the A64 would gain more performance due to its greater performance/clock scaling.. but that is another subject.

Dazz
01-24-04, 01:10 PM
Yeah but the P4 gets very big peformance boosts from extra cache, look up the Willamate Vs NorthWood. The look at the PR of the Athlons, the diffrence from 512k & 1024k = 1~3% diffrence while it alot bigger on the Pentium 4 due to the longer pipelines which will use more cache.

Nutty
01-24-04, 04:47 PM
True, proving its already bottlenecked by slow main RAM. Pushing the core Ghz faster will only increase this bottleneck.

Huge amounts of cache are a solution, but its pricey.

Nephilim
01-24-04, 08:01 PM
Originally posted by Viral
Exactly, it's all relative.. You get a higher overclock, so you think you are getting better performance, but really you are getting nothing extra at all.


Actually you are, but you probably don't want it......heat.

I had always thought that, as a general rule, longer pipes were Bad Things. I mean, what happens when you send an instruction down the pipe only to have the error checking pick something up near the end so that it has to send that instructions down the long-ass pipe again. Multiply that a few million times. Bigger cache or not, you'd still be buggered. Granted, this is worse case senario, but even so, wouldn't shorter pipes mean more instructions per clock cycle?

Nutty
01-24-04, 08:15 PM
Indeed, but at lower clock speeds. ;)

I read an article about increasing pipeline depths in order to improve performance, and they concluded the optimal was around 50 stages for a P4. Tho it was all theoretical. It said something like;

We concluded that a 100% clock speed increase with X extra pipes can yield 30% to 90% extra performance.

I was like thinking, so if they mess it up completely like NV3X architecture, we're looking at 30% performance increase at 6ghz? Or double performance at 6Ghz, if they nail it perfectly..

The whole point of that article was to say, "Yeah we know chips are gonna suck at 1st, but when Joe Public seens a 4Ghz sticker on the box, they'll be like "Wooa Yeah, gotta get me one of those bad boys!" "

Nephilim
01-24-04, 09:06 PM
*laugh*

While they're at it, I may have a bridge they'd be interested in buying.

StealthHawk
01-24-04, 11:55 PM
Originally posted by Dazz
Yeah but the P4 gets very big peformance boosts from extra cache, look up the Willamate Vs NorthWood. The look at the PR of the Athlons, the diffrence from 512k & 1024k = 1~3% diffrence while it alot bigger on the Pentium 4 due to the longer pipelines which will use more cache.

Pentium4's cache is better(more efficient) than Athlon's cache last I checked so that contributes to the performance discrepancy as well.

Viral
01-26-04, 07:31 AM
http://www.overclockers.com/articles940/

Some comentary on the situation..

If you have a CPU that can run at 4GHz, and extending the pipeline means you can reach 5 or 5.5GHz, that's one thing, and probably a good thing. On the other hand, if you have a CPU that is supposed to do 4GHz, but can only do 3GHz, and this is the only way you can make it to go to 4GHz, that's not a good thing, in fact, it's pretty bad.

The real test will come when the socket 478 and socket Ts get overclocked. If what I've said is true, these chips will probably get into the low 4GHz range. However, it's going to be a devalued 4GHz which will hardly be any better than the 3.6GHz or so a good Northwood can do today.

I had said a little while back I was leaning towards buying a socket T system as the least bad upgrade. After this information, not any more.

They seem to share my oppinion.. which is a change from the regular stuff i read from overclockers:rolleyes: Though i don't agree with the entire article.. so i guess the world hasn't turned on itself just yet.;)

SH0DAN
01-26-04, 05:03 PM
Originally posted by bkswaney
For us overclockers this is great.
Longer pipes = higher clocks. :D
Even if it comes out at 3.4 if it can hit 4+ghz I'll be most happy. ;)


we need to see how much heat the pressy puts out first....i'm looking
for another 1.6A Malay like my two that'll both do 2.7 to 2.9 on air.i
need to see HUGE jumps in performance over what i have currently before i spend any dough,on any new chips.