Intels Rival for the Hammer!!!

[SSXeon5]

His name is:

PRESCOTT

-Socket mPGA478
-Pentium 4 .09um 4GHz-6Ghz
-800Mhz FSB (200 quad pumped)
-1MB L2 Full speed cache
-32-bit ALUs Full speed
-Dual channel DDR-II 400Mhz
-Hyperthreading (More advanced Ver.)
-Advanced hardware data prefetch

Dont tell me Hammer was the only game in town !!

SSXeon

 

[dman]

Still 32bit though, eh?

 

[meson2000]

To bad it is going to be out a full year after Hammer Debuts......
Q4 2002 vs. Q4 2003.......sorry Intel, no dice.....

 

[MisterBombastic]

And too bad 'Prescott' is a rubbish name. Why not call it 'Blair'?!

 

[BlvdKing]

Maybe Prescott will have "Yamhill" technology enabled? We'll see...

 

[Colt45]

uh.. its still 32 bit.

besides, you usually pay through the teeth when youre buying a intel SERVER processor, and its still only 32 bit.

and who is to say it wont have 45,000 pipelines so that it has a real low IPC?
(thats how P4 seems to be going)

i'd like to see a .09um P6 based chip with 1mb L2, 200 FSB, DDR-II and some crazy high clockspeed. basically P6 with some of the P7 architecture advances.

i actually heard they're making one for laptops, but why not desktops? (cant think of the name right now)

 

[christoph83]

To bad it is going to be out a full year after Hammer Debuts......

Link? Prescott chipsets are coming out in the first half of 2003 . Wouldnt that mean that the cpu would come out soon after?

and who is to say it wont have 45,000 pipelines so that it has a real low IPC?

Why would it have 45,000 pipelines? This a P4 were talking about. The same P4 at .18 and .13 processes. The prescott is just a P4 die shrink with newer technology. Hyperthreading is sopposed to be included with the prescott as well.

besides, you usually pay through the teeth when youre buying a intel SERVER processor, and its still only 32 bit.

Who says this is a server processor?

 

[Colt45]

<< Who says this is a server processor? >>



my bad, I thought prescott was the next socket 602 xeon.

 

[imgod2u]

I don't know why the focus is on this 64-bit/32-bit thing. The "bitness" of the processor only refers to the memory address and the ALU. FPU has been 64-bit for a while now (double precision). I don't know about you but I cannot imagine a common day application that could use 64-bit ALU's. As for the memory address, the 4-bit extension means 64-bit isn't really needed. Will it help in the scientific world where 64-bit ALU's are useful? Sure. But that's a niche market and is already in full use of modern 64-bit processors such as Spark and Power.

 

[SSXeon5]

I said Socket mPGA478 which is what it is at this point, It might be called the Pentium 5 or something. The xeon ver of this will have 3MB+ onboard L2 cache and will be dual and quad compadible. And It will come out (Q1'03-Q2'03) ..... I cant wait for it ......

SSXeon

 

[AGodspeed]

Link? Prescott chipsets are coming out in the first half of 2003 . Wouldnt that mean that the cpu would come out soon after?

Prescott is due in the second half of 2003, not the first half. Linkage.

PRESCOTT

-Socket mPGA478
-Pentium 4 .09um 4GHz-6Ghz
-800Mhz FSB (200 quad pumped)
-1MB L2 Full speed cache
-32-bit ALUs Full speed
-Dual channel DDR-II 400Mhz
-Hyperthreading (More advanced Ver.)
-Advanced hardware data prefetch.

According to Intel Prescott will look like the following:

1. .09-micron die shrink from .13-micron "Northwood" core.
2. Additional unspecified micro-architectural enhancements.
3. Hyper-Threading Enabled.

.09-micron P4's by themselves will yield more frequency increases and will run cooler at the same clock speed compared to .13-micron P4's. Performance will hardly be any different between, lets say, a 3.5GHz Northwood (.13-micron) and 3.5GHz Prescott (.09-micron) if all Prescott will be is a die shrink, but it's obviously going to be more than that.

I haven't once heard of a more "advanced version" of Hyper-Threading before. Do you have any info on this that we might be able to look at.

From reading the net, it seems like HT might or might not increase performance for Joe Average-type apps. Remember that HyperTreading is highly dependent on what application you're using, and in fact some apps actually will perform worse when HyperThreading is enabled. Prelimary results from web sites like AnandTech and others seem to indicate that HT will mostly gain an advantage in workstation/low-end server type apps, not Joe Average OEM situations (you know what I'm talking about, a little gaming, some web surfing, DVD's, mp3's; those types of Joe Average apps).

An 800MHz FSB Prescott is speculation at this point (Anand mentioned this 800MHz figure in one of his IDF articles in late February). Since Anand's speculation though, I've seen 666MHz FSB a few times on "unofficial" Intel roadmaps (166.666... * 4 = 666.666...MHz FSB).

32-bit ALU's is also speculation. To be honest I don't know how much of an improvement 32-bit ALU's will yield over 16-bit ALU's.

1MB of L2 cache is certainly possible, and actually pretty likely IMHO. Intel has lots and lots of capacity. Even though they'll get fewer CPU's per wafer than if they were to go with a 512K L2 cache, I bet they'll still be able to supply enough Pentium 4 processors to their OEM and corporate partners with Prescott (although so much can happen a year from now...).

If Prescott actually does actually debut with DDR-II support (meaning there would have to be sufficient supplies of DDR-II on the market) then I can guarantee you that Prescott will be delayed to sometime in 2004. DDR-II isn't expected to yield any type of major (or even minor) production in 2003. Prescott will likely just support DDR333 if it debuts sometime between July and December (which is considered "2nd Half" of 2003).

 

[christoph83]

Prescott is due in the second half of 2003, not the first half. Linkage.

That link lead to intel's developer's forum but I couldnt find a roadmap. Anyway the only point I was making is that according to your roadmap in this forum prescott chipsets are coming out in the first half of 2003, which usually means the cpu should be out soon after. But all of this is really speculation at this point.

From reading the net, it seems like HT might or might not increase performance for Joe Average-type apps. Remember that HyperTreading is highly dependent on what application you're using, and in fact some apps actually will perform worse when HyperThreading is enabled. Prelimary results from web sites like AnandTech and others seem to indicate that HT will mostly gain an advantage in workstation/low-end server type apps, not Joe Average OEM situations (you know what I'm talking about, a little gaming, some web surfing, DVD's, mp3's; those types of Joe Average apps).

You could say the same about the 64 bit instructions in the new clawhammer. And considering this technology is brand new, once it matures it may turn out to be a really nice feature. If we are infact 14 months away from a prescott release, thats a long time for this technology to mature. Although Average Joe still probably wont see a difference from hyperthreading or 64 bit instructions.

 

[SSXeon5]

<< Prescott is due in the second half of 2003, not the first half. Linkage.

That link lead to intel's developer's forum but I couldnt find a roadmap. Anyway the only point I was making is that according to your roadmap in this forum prescott chipsets are coming out in the first half of 2003, which usually means the cpu should be out soon after. But all of this is really speculation at this point.

From reading the net, it seems like HT might or might not increase performance for Joe Average-type apps. Remember that HyperTreading is highly dependent on what application you're using, and in fact some apps actually will perform worse when HyperThreading is enabled. Prelimary results from web sites like AnandTech and others seem to indicate that HT will mostly gain an advantage in workstation/low-end server type apps, not Joe Average OEM situations (you know what I'm talking about, a little gaming, some web surfing, DVD's, mp3's; those types of Joe Average apps).

You could say the same about the 64 bit instructions in the new clawhammer. And considering this technology is brand new, once it matures it may turn out to be a really nice feature. If we are infact 14 months away from a prescott release, thats a long time for this technology to mature. Although Average Joe still probably wont see a difference from hyperthreading or 64 bit instructions. >>



Your right it Will come out the first half of 2003 .... even quicker seeing how Intel sped up the new fabs and 10% smaller 13micron cores means faster speeds then predicted. The 3.06Ghz northwood will most defiantly hit in Q4 2002 and prescott will start at 3.5-4Ghz. Someone said that a 3.5Ghz northwood will be the same performance as the 3.5GHz prescott .... this is false. The prescott will get a overhauled core .... 2x the cache of northwood, more efficient use of hyperthreading, 800Mhz fsb, higher IPC, and DDR-II. Why DDR2 well JEDEC already approved the specs for DDR2 in the beginning of 2002 and I read that they will be mass produced in Q1-Q2 2003 ..... also prescotts launch .... hmmm. PC-3200 (DDR400) was not approved by JEDEC so DDR2 is the way to go at this point. Im behind intel 100% (check the sig ) And I really think at this point prescott can surly outperform the hammer. Intel even has there own version of x86-64 (yamhill) just in case cant wait for 2003!

SSXeon

 

[AGodspeed]

That link lead to intel's developer's forum but I couldnt find a roadmap. Anyway the only point I was making is that according to your roadmap in this forum prescott chipsets are coming out in the first half of 2003, which usually means the cpu should be out soon after. But all of this is really speculation at this point.

AnandTech's IDF Coverage

- 0.09-micron (90nm)
- additional micro-architectural enhancements
- Hyper-Threading on desktops
- shipping in second half 2003

Some people are saying summer of 2003 as Prescott's launch. Who knows...

You could say the same about the 64 bit instructions in the new clawhammer...

How can you say the same thing about x86-64? x86-64 isn't going to reduce performance in any apps, it simply enables the option of running 64-bit apps and 64-bit OS's while still being able to run all your current 32-bit apps and 32-bit OS's. Sure Joe Average isn't really going to be running 64-bit apps or OS's, but that doesn't matter at all; ClawHammer will be able to run 32-bit apps and OS's faster than the Athlon XP.

SSXeon5 said the following:

The 3.06Ghz northwood will most defiantly hit in Q4 2002...

The prescott will get a overhauled core....more efficient use of hyperthreading, 800Mhz fsb, and DDR-II.

Why DDR2 well JEDEC already approved the specs for DDR2 in the beginning of 2002 and I read that they will be mass produced in Q1-Q2 2003

SSXeon5, you need to site your sources and not just spout this kind of stuff.

A 3.06GHz Northwood in Q4 2002 is possible, but not likely given that Intel would have to skip a couple speed grades and wouldn't be able to bank on lower speed grade Pentium 4's.

What does a "more efficient use of hyperthreading" mean? And where did you hear about this. Are you making stuff up?

And like I said before, an 800MHz FSB is just speculation at this point. Other web sites have mentioned a 667MHz FSB. Do you have any sources to back up your claim that Prescott will "defiantly" come with an 800MHz FSB?

And you're quite mistaken about DDR-II's projected production timetable. The vast majority of DRAM manufacturers have said that DDR333 (DDR-I) won't reach mass production/market acceptance until Q1 of 2003. Later on in 2003 DRAM makers will start producing DDR-II modules, but it won't reach the mainstream (mass production) until 2004. Intel wouldn't release a DDR-II supporting CPU/chipset if there was no DDR-II on the market.

Here are a couple articles on DDR-II articles for you to read:

EETimes

SiliconStrategies

 

[imgod2u]

When Intel cut the number of FPU with the P4 design, they did it because they knew the pipeline would not be able to fill up any more. With hyperthreading, more execution units would be needed in order to process all the instructions from the two threads. As seen from the Sandra scores, the double-pumped ALU (effectly 4 ALU) helps a lot when Hyperthreading is enabled. I would expect that with Hyperthreading enabled on the desktop processor, Intel would do something similar with the FPU. The Sandra scores for the FPU with HT enabled didn't show much of an increase at all.

 

[MadRat]

Sandra is a synthetic benchmark. If the program isn't rewritten then it will not be aware of how to use HT to improve the scores.

 

[DivideBYZero]

SSXeon> With regard to Yamhill this is in response to Hammer, so Intel have a long way to go. AMD have shown final silicon, intel are still jamming in the back rooms with Yamhill.

Prescot can suck AMD's balls, High Ghz and no punch. All revs and no Torque.

Yamhill is the worst kind of VapourWare and the Prescott will be yet another P2-P3 style transition with little added and a lot taken away(from your wallet).

Yes, this is Intel fan-boy flamebait...

 

[Degenerate]

My $2.

It seems that from what i have read, P4 should hit 3 Ghz by this years end. However, what ever it hits, (mabey 2.8-9), once the Hammer is released, you can be sure that intel will release P4 to closely match the perfomance of hammer. This is provided thatthe hammer is faster than Xp's.

BTW, what clock speeds will the hammer be released? i am guessing about 2.5Ghz?

 

[spanner]

How much of a clock speed increase do you think a .13 to .09 die shrink will give? A ghz at the most if you ask me. Plus it is like a year away from now. I am also assuming these things will need new motherboards too. As for DDR-II support, Amd is already planning a new hammer once DDR-II becomes mainstream, so thats no advantage. Clawhammers already have the ability to go to 1mb cache if needed and Amd will eventually be going for a .09m process. To be honest I seriously hope Intel can do a tad better then this else the hammer will crush them.

 

[SSXeon5]

<< According to Intel Prescott will look like the following:

1. .09-micron die shrink from .13-micron "Northwood" core.
2. Additional unspecified micro-architectural enhancements.
3. Hyper-Threading Enabled.

.09-micron P4's by themselves will yield more frequency increases and will run cooler at the same clock speed compared to .13-micron P4's. Performance will hardly be any different between, lets say, a 3.5GHz Northwood (.13-micron) and 3.5GHz Prescott (.09-micron) if all Prescott will be is a die shrink, but it's obviously going to be more than that.

I haven't once heard of a more "advanced version" of Hyper-Threading before. Do you have any info on this that we might be able to look at.

From reading the net, it seems like HT might or might not increase performance for Joe Average-type apps. Remember that HyperTreading is highly dependent on what application you're using, and in fact some apps actually will perform worse when HyperThreading is enabled. Prelimary results from web sites like AnandTech and others seem to indicate that HT will mostly gain an advantage in workstation/low-end server type apps, not Joe Average OEM situations (you know what I'm talking about, a little gaming, some web surfing, DVD's, mp3's; those types of Joe Average apps).

An 800MHz FSB Prescott is speculation at this point (Anand mentioned this 800MHz figure in one of his IDF articles in late February). Since Anand's speculation though, I've seen 666MHz FSB a few times on "unofficial" Intel roadmaps (166.666... * 4 = 666.666...MHz FSB).

32-bit ALU's is also speculation. To be honest I don't know how much of an improvement 32-bit ALU's will yield over 16-bit ALU's.

1MB of L2 cache is certainly possible, and actually pretty likely IMHO. Intel has lots and lots of capacity. Even though they'll get fewer CPU's per wafer than if they were to go with a 512K L2 cache, I bet they'll still be able to supply enough Pentium 4 processors to their OEM and corporate partners with Prescott (although so much can happen a year from now...).

If Prescott actually does actually debut with DDR-II support (meaning there would have to be sufficient supplies of DDR-II on the market) then I can guarantee you that Prescott will be delayed to sometime in 2004. DDR-II isn't expected to yield any type of major (or even minor) production in 2003. Prescott will likely just support DDR333 if it debuts sometime between July and December (which is considered "2nd Half" of 2003). >>



I wont flame yeah cuz texmaster said your a good guy Yea you have a good point, pretty much every future thing is speculation at this point, all the hammer info ect ... remember the 1MB L2 cahce Thunderbird LMFAO .... that never happened, they scraped that idea. I still dont think that the hammer will arrive on time, given AMD's history of delays, but it still looks like a good idea im my head ..... and as for the DDR-II ... I still believe intel can get is with prescotts launch .... they will prob just bundle it like they did with rambus ....




<< How much of a clock speed increase do you think a .13 to .09 die shrink will give? A ghz at the most if you ask me. Plus it is like a year away from now. I am also assuming these things will need new motherboards too. As for DDR-II support, Amd is already planning a new hammer once DDR-II becomes mainstream, so thats no advantage. Clawhammers already have the ability to go to 1mb cache if needed and Amd will eventually be going for a .09m process. To be honest I seriously hope Intel can do a tad better then this else the hammer will crush them. >>



The prescott will use the mPGA478 socket from what I have read and the speed will be from 4Ghz-6Ghz because the northwood core will hit about 3.5GHz+ in Q2 2003. Intel like I said has been speeding up there fabs and now have the 300mm wafers to use and gives the new northwoods 10% smaller core then the first. The 2.4Ghz Xeons will have the new 10% smaller core and the new B's will also feature this. As for the hammer ... the clawhammer (athlon) will have 256k-512k L2 cahce and the Sledgehammer (opteron) will have up to 1MB L2 ..... yet The Prescott will feature 1MB L2 and the Xeon ver have up to 3MB+ L2 cahce. The thing is that the sledgehammer will have to face Madison (.13um IA-64) and it will feature 500million transistors, 6-8MB L2 cache, and pretty much double McKinley. The hammer looks pretty good but like I said before I need more then just a ball bouncing in linux to make me impressed

SSXeon

 

[spanner]

"The thing is that the sledgehammer will have to face Madison (.13um IA-64) and it will feature 500million transistors, 6-8MB L2 cache, and pretty much double McKinley"

Not sure but from what I have read, the 500million transistor monster is targeting a market somewhat above what the hammer is going for. Intel really has no hammer equivalent that can be up and running in time to squash the hammer. I mean a 3.5Ghz northwood will not have the same advantages the 1.6A has, i.e overclockability, low power consumption etc. Intels plan is more likely to be along the lines of a pricewar, taking a loss to undercut the hammers untill they can get their act together, atleast that would be the honest thing to do, but I believe they will probably continue their little smear campaign thats been going around for a while.

 

[Sid03]

<< How much of a clock speed increase do you think a .13 to .09 die shrink will give? A ghz at the most if you ask me. >>

and what do you base your opinion on?



<< I am also assuming these things will need new motherboards too. >>

again, why are you assuming this? besides, if needing a new board is such an issue for you, then i guess you have problems with the hammer, too.



<< Amd will eventually be going for a .09m process. >>

yep, and amd will eventually be going to a .13 micron process too. a full year (or more) after intel. what's your point?



<< To be honest I seriously hope Intel can do a tad better then this else the hammer will crush them. >>

to be honest, you have no basis for your opinions other than to spread fud, as you always do. if you do a search of your posts, you'll find that they are nothing but a bunch of amd zealotry.

with comments like: "As much as I despise intel..." i'm not sure how one could take your comments as unbiased.

 

[dexvx]

What the freak is up with Itanium/Hammer comparisons?

They're aimed at different market places, the only markets they will actually compete in is the 1-8 processor server/workstation. Other than that, Itanium is aimed at hundred+ processor servers due to its high scalability. Even the freaking Merced had a couple of 1000+ processor servers made under it by IBM and HP.

Everything else is hypothetical performance on circumstantial evidence for the most part.

 

[spanner]

hmm lets see. the 1 ghz increase is based on the increase from .18 to .13 micron. Seems I wasn't too far off either from all the other info i.e nothwood leaves off at 3.5 ghz and prescott picks up at 4ghz, I mean its not like they are going to debut ay 6ghz right off the bat???

New motherboards for DDR-II support ofcourse. New motherboard requirements should really come with huge processor design changes, not minor ones.

My point about the die size is that Amd can do to their processors pretty much whatever intel does so if Amd starts off ahead with the hammer then they can pretty much take another leap every time intel does. Sure they are a year behine with the .13 micron jump but they are still not far behind intels processors.

As for being biased I never said I wasn't, i just want a little competition. How about posting your opinions rather then attacking mine. As we all know all of this is based on speculation so take my words as speculation.

 

[unclejimbo]

<< His name is:

PRESCOTT

-Socket mPGA478
-Pentium 4 .09um 4GHz-6Ghz
-800Mhz FSB (200 quad pumped)
-1MB L2 Full speed cache
-32-bit ALUs Full speed
-Dual channel DDR-II 400Mhz
-Hyperthreading (More advanced Ver.)
-Advanced hardware data prefetch

Dont tell me Hammer was the only game in town !!

SSXeon >>




I first must say, the Prescott is NOT a Pentium 4 it IS a Pentium 5, it will enter the market sometime in January at speeds of aorund 4Ghz, and I doubt it would use Socket478.


-- unclejimbo