I think everyone here is overestimating SSE2 penetration, and doubly overestimating the attraction of SSE2 for AMD. SSE2 is tailored tot he P4 and is not nearly as useful on a non-P4 CPU. AMD will get it when/if everything in the performance software arenas supports it.

Originally Posted by GiGNiC
with a little luck, maybe Barton will do SSE2 aswell

Originally Posted by GiGNiC
K8 will do SSE2

It's all gonna be about SSE2. The most important aspect of the 845 chipset (soth SDR and DDR) is that it will give the P4 and thus SSE2 much greater market penetration. Then, programmers will start writing/compiling for SSE2. This should give all P4's a nice little boost.
Not to mention that the die-size shrink is not only about higher clocks and lower heat, but also lower production costs. This is why you see Northwoods for nearly the same price as .18 micron P4's. The die will be smaller even with the extra cache. Hopefully Intel will be able to be more profitable and be able to sell stuff cheaper.
What is the next AMD proc that is going to use SSE2?

[qou may find yourself wanting an Intel CPU agai/q]
lol, I think not,
Even if it does outperform AMD's chips, what will be the price/performance ratio??? I, like a lot of other people, am not willing to pay twice the amount for a 5% increase in performance.

Here's the latest news from HARDOCP.COM:

"We have been mentioning to you that the Northwood Pentium4 CPU will wake a few of you up when it comes out. Here are some OC results that will for the most part not be that unusual if our theories are correct.

First off, Azzkicker from the [H]ardForum, sent in this Japanese page that shows a near 3GHz CPU OC. Then VRZone chimes in with there 2.8GHz OC of the 2.2GHz version of the P4. If you check our VapoChill review you will find an "old" Willamette core P4 doing 2.5Ghz and posting some of the highest benchmarks we have ever seen. Give those CPUs a chance to come down a little in price and by summer you may find yourself wanting an Intel CPU again."

The VRZone 2.8GHz OC is available at the following url:

http://vr-zone.com/cgi-bin/vb/showthread.php?s=d4d7f92361f2169e2aa0...

#21, how can you make that assumption with little or no data? Making the assumption that since one Pentium 4 will do 3000 mhz, all will do so or even a majority will do so is very bad science.

#20--

The current P4 is a .18u and is twice the die size of the new .13u "Northwood" P4 (2.2GHz cpu). The new .13u "Northwood" also runs at a fraction of the default voltage and produces a fraction of the heat. The 2.2GHz clocked at 3.0GHz probably runs no hotter than a current 2.0GHz P4 clocked at 2.2 or 2.3GHz.

#19 -- it's not hard to believe, that's generally the overclocking consensus for the Pentium 4 -- 300 mhz above default shipping. It's a greater leap of faith to assume 800+ mhz overclock.

All sorts of sites (Tomshardware, Anandtech included, as you'll see early next week) have clocked retail 2.2's up to 2.5ghz and 2.6ghz with the stock Intel heatsink and default voltage. Why is it so hard to believe that replacing the Intel heatsink with a superior Swiftech, and increasing the voltage can't yield even better results?

That said, what I am really looking forward to are the benchmarks of the Prestonia with SMT next week. Originally it was thought that this cpu wasn't going to be out until February or March, but a number of vendors have received shipments of these cpus in the past week. There are threads on 2cpu.com where members have ordered these cpus for their Xeon systems (Supermicro motherboards now support SMT). The NDA on this cpu also expires next week, or so people are saying.

Here's one vendor that has Prestonia (P4 w/ 512Kb cache and SMT enabled) in stock that is shipping them as of today--erm yesterday--via Fedex:

http://8anet.com/listprod.ihtml?parent=24

Their pricing is $348 for the 1.8GHz Prestonia with 512Kb cache and SMT, and $585 for the 2.0GHz version. They expect oem stock of the 2.2GHz version next week. By comparison, they charge $299 for the Athlon MP 1800+ and $242 for the Athlon XP 1800+. They clearly don't have the best prices, so it should be available for less from others next week. Another vendor in the UK charges LESS for the 2.0GHz Prestonia than it does for the 2.0GHz Northwood, and LESS for the 1.8GHz Prestonia (unlike Northwood, Prestonia will also be available in a 1.8GHz version) than it does for the Athlon MP 1800+.

voltage is cranked up pretty high, if you consider northwood is supposed to be at (1.6 volts is it? 1.4? I don't recall correctly)

I just found the bus speeds and Sandra benches off. H.Oda authentication would make me a 100% believer.

Originally Posted by Trident Troll
Intel had one of these running at 3Ghz several months ago.
Why is this one so hard to believe?

Right. I'm saying he needs to get H.Oda to verify it.

Forge, I bring your attention to the WCPUID window in the screenshots

If he runs WCPUID, it's verifiable. H.Oda can verify the authenticity of the checksum in the lower right corner of the window. This guy needs to run WCPUID and get it verified before I care.

Originally Posted by GiGNiC
ah well, to spare you all the trouble:

[quote]1.Intel Pentium4 Northwood 2.2GES(Vcore:1.9V)
2.Asus P4S333 Rev 1.01(Bios:1002)
3.KingMax DDR333 CL2.5 256MB DDRRAM*2(Vddr:2.9V)
4.Swiftech MCX-462 for Socket478 w/Delta FFB0812EHE Fa/quote]

Originally Posted by GiGNiC
just read the actual article

it says in plain alfabet which components are used

[quote]4.Swiftech MCX-462 for Socket478 w/Delta FFB0812EHE Fa/quote]

the link isn't dead

I saw liquid in the article somewhere. Maybe they used liquid nitrogen?

The link to the benchmarks is dead.

dB

Anyone run this through babel fish to find out what they used to get it this high?

If this is true, it is a bad omen for AMD. Bring on .13 micron.

Hmm, I dunno. Why do you think Sandra pics look edited? This dude uses Chinese Windows, so the default font is obviously different from English Windows (ask anyone who uses Far East versions of Windows... it's common). Hence a serif font instead of the usual sans-serif font in Sandra. Also, the bar's gradation looks 'simplified' because it was posted as a GIF file (which, of course, can only display 256 colours, resulting in less natural gradation).

Another point... a same class of system can have RAM bench points differ by hundreds of points... by some tweak or otherwise. And both Sandra and WCPUID have been known to misreport FSB of 'yet unreleased at the compile time' systems (Sandra being incorrect more often).

In conclusion, I am thinking that the pics are more of a genuine thing. They can be a very refined fake, but I don't see any visibly unusual things about them. Please do point out anything I've missed.

[i]amddiyoc has overclocked the Pentium 4 2.2GHz to 3.0GHz on the Asus P4S333. [/i]

Yeah, but he only held it for a few seconds before his house burst into flames!!! (joke)

I agree with #1 & #2...funny quere mmm hummm

Originally Posted by GiGNiC
also, on wcpuid there\'s a 600Mhz FSB, but in sandra it\'s 748Mhz
clockspeeds are at least a constant 3Ghz

an other observation, the multiplier is adjustable ? the d00d has an engineeringsample then ? (sandra has a 16x multiplier, wcpuid has 19x and 20x)

something totally different: notice the estimated PR-ratings: 3300 :) Hammer should be at 3400+ in the same timeframe, nice to see

to me it looks strange, too.

besides, those benches don´t tell you how fast it really is.

Damage, please take one of those babes and show us how it performs!

Originally Posted by NeonKrist
Do the Sandra benchmarks look fake to anyone else? Can\'t tell from the WCPUID screens, but Sandra looks edited. You\'d also expect the RAM to be at least a couple hundred points faster with the 600mhz FSB. And finally, it seems just a little bit strange that the FSB and CPU speeds are so perfectly round.